home *** CD-ROM | disk | FTP | other *** search
Wrap
Text File | 1991-12-22 | 118.8 KB | 4,586 lines
.rs .\" Troff code generated by TPS Convert from ITU Original Files .\" Not Copyright (~c) 1991 .\" .\" Assumes tbl, eqn, MS macros, and lots of luck. .TA 1c 2c 3c 4c 5c 6c 7c 8c .ds CH .ds CF .EQ delim @@ .EN .nr LL 40.5P .nr ll 40.5P .nr HM 3P .nr FM 6P .nr PO 4P .nr PD 9p .po 4P .rs \v'|.5i' .LP \fBMONTAGE:\ \ \fR Fin de la Rec. X.134 en t\*\|ete de cette page .sp 2P .LP \v'30P' \fBRecommendation\ X.135\fR .RT .sp 2P .ce 1000 \fBSPEED\ OF\ SERVICE\ (DELAY\ AND\ THROUGHPUT)\ PERFORMANCE\ VALUES\fR .EF '% Fascicle\ VIII.3\ \(em\ Rec.\ X.135'' .OF '''Fascicle\ VIII.3\ \(em\ Rec.\ X.135 %' .ce 0 .ce 1000 \fBFOR\ PUBLIC\ DATA\ NETWORKS\ WHEN\ PROVIDING\fR .ce 0 .sp 1P .ce 1000 \fBINTERNATIONAL\ PACKET\(hySWITCHED\ SERVICES\fR .ce 0 .sp 1P .ce 1000 \fI(Malaga\(hyTorremolinos, 1984; amended at Melbourne, 1988)\fR .sp 9p .RT .ce 0 .sp 1P .LP The\ CCITT, .sp 1P .RT .sp 1P .LP \fIconsidering\fR .sp 9p .RT .PP (a) that Recommendation X.1 specifies the international user classes of service in public data networks; .PP (b) that Recommendation X.2 specifies the international data transmission services and optional user facilities in public data networks; .PP (c) that Recommendation X.25 specifies the DTE/DCE interface for packet mode terminals connected to public data networks by dedicated circuit; .PP (d) that Recommendation X.75 specifies the packet switched signalling system between public networks providing data transmission services; .PP (e) that Recommendation X.323 specifies general arrangements for interworking between packet\(hyswitched public data networks; .bp .PP ( f ) that Recommendation X.96 specifies call progress signals in public data networks; .PP (g) that Recommendation X.110 specifies the international routing principles and routing plan for public data networks; .PP (h) that Recommendation X.213 defines the OSI Network Layer service; .PP (i) that Recommendation X.140 defines general quality of service parameters for communication via public data networks; .PP ( j ) that Recommendation X.134 specifies portion boundaries and packet layer reference events for defining packet\(hyswitched performance parameters; .PP (k) that Recommendation X.136 specifies accuracy and dependability (including blocking) performance values for public data networks when providing international packet\(hyswitched service; .PP (l) that Recommendation X.137 specifies availability performance values for public data networks when providing international packet\(hyswitched service, .sp 1P .LP \fIunanimously declares\fR .sp 9p .RT .PP (1) that the speed of service parameters defined in this Recommendation shall be used in the planning and operation of international packet\(hyswitched data communication services provided in accordance with Recommendations\ X.25 and\ X.75; .PP (2) that in such services, the performance values specified in this Recommendation shall be taken as worst\(hycase limits under the conditions specified herein. .sp 2P .LP \fB1\fR \fBIntroduction\fR .sp 1P .RT .PP 1.1 This Recommendation is the second in a series of four CCITT Recommendations (X.134\(hyX.137) that define performance parameters and values for international packet\(hyswitched data communication services. Figure\ 1/X.135 illustrates the scope of these four Recommendations and the relationships among them. .sp 9p .RT .PP 1.2 Recommendation X.134 divides a virtual connection into basic sections whose boundaries are associated with X.25 and X.75 interfaces; defines particular collections of basic sections, called virtual connection portions, for which performance values will be specified; and defines a set of packet layer reference events (PEs) which provide a basis for performance parameter definition. The basic sections consist of network sections and circuit sections. They are delimited, in each case, by physical data terminal equipment (DTE) or data switching equipment (DSE) interfaces. Virtual connection portions are identified either as national portions or international portions. Each PE is defined to occur when a packet crossing a section boundary changes the state of the packet layer interface. .sp 9p .RT .PP 1.3 For comparability and completeness, packet\(hyswitched network performance is considered in the context of the 3\ \(mu\ 3 performance matrix defined in Recommendation\ X.140. Three protocol\(hyindependent data communication functions are identified in the matrix: access, user information transfer, and disengagement. These general functions correspond to call set\(hyup, data (and interrupt) transfer, and call clearing in packet\(hyswitched virtual call services conforming to the X.25 and X.75 Recommendations. Each function is considered with respect to three general performance concerns (or \*Qperformance criteria\*U): speed, accuracy, and dependability. These express, respectively, the delay or rate, degree of correctness, and degree of certainty with which the function is performed. .sp 9p .RT .PP 1.4 This Recommendation defines protocol\(hyspecific speed of service parameters and values associated with each of the three data communication functions. Recommendation\ X.136 defines protocol\(hyspecific accuracy and dependability parameters and values associated with each function. The Recommendation\ X.135 and Recommendation\ X.136 parameters are called \*Qprimary parameters\*U to emphasize their direct derivation from packet layer reference events. .bp .sp 9p .RT .LP .rs .sp 40P .ad r \fBFigure 1/X.135, (N), p.1\fR .sp 1P .RT .ad b .RT .PP 1.5 An associated two\(hystate model provides a basis for describing overall service availability. A specified availability function compares the values for a subset of the primary parameters with corresponding outage thresholds to classify the service as \*Qavailable\*U (no service outage) or \*Qunavailable\*U (service outage) during scheduled service time. Recommendation\ X.137 specifies the availability function and defines the availability parameters and values that characterize the resulting binary random process. .sp 9p .RT .PP 1.6 Four speed of service parameters are defined in this Recommendation: one access parameter (call set\(hyup delay), two user information transfer parameters (data packet transfer delay and throughput capacity), and one disengagement parameter (clear indication delay). Each parameter can be applied to any basic section or portion of a virtual connection. This generality makes the parameters useful in performance allocation and concatenation. .bp .sp 9p .RT .PP 1.7 This Recommendation specifies delay and throughput values for national portions and international portions of two types (Table\ 1/X.135). Performance values for data terminal equipment are not specified, but the parameters defined in this Recommendation may be employed in such specification to assist users in establishing quantitative relationships between network performance and quality of service (see Recommendation\ X.140). .sp 9p .RT .ce \fBH.T. [T1.135]\fR .ce TABLE\ 1/X.135 .ce \fBVirtual connection portion types for which\fR .ce \fBperfomance values are specified\fR .ce \|\ua\d\u)\d .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(60p) | cw(120p) . Portion type Typical characteristics _ .T& lw(60p) | lw(120p) . National A T{ Terrestrial connection via an access network section T} _ .T& lw(60p) | lw(120p) . National B T{ Connection via an access network section with one satellite circuit; or via an access network section and one or more transit network sections T} _ .T& lw(60p) | lw(120p) . International A T{ Connection via a direct terrestrial internetwork circuit section T} _ .T& lw(60p) | lw(120p) . International B T{ Connection via two satellite circuits and one transit network section; or via one satellite circuit and two or more transit network sections \ua\d\u)\d The values specified for Type B portions also apply to virtual connection portions not explicitly identified as Type A or Type B. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 1/X.135 [T1.135] p.\fR .sp 1P .RT .ad b .RT .PP 1.8 Worst\(hycase mean and 95% probability values for call set\(hyup delay, data packet transfer delay, throughput capacity, and clear indication delay are specified for each virtual connection portion type identified in Table\ 1/X.135. The term \*Qworst case\*U means that these values should be met during the normal busy hour in the worst\(hyperforming virtual connection portion used in providing international packet\(hyswitched service. The performance of a virtual connection portion will normally be much better than the worst\(hycase values specified in this Recommendation. .FS Supplement\ 1 presents delay and throughput values measured on particular connections at particular times and is for illustrative purposes only. .FE Design objectives that take into account more demanding user applications and network performance and connectivity enhancements are for further study. .sp 9p .RT .PP Numerical methods for combining individual portion performance values to estimate end\(hyto\(hyend performance are also provided in this Recommendation. DTE to DTE values for two particular hypothetical reference connections are derived using these methods in Annex\ C. .sp 2P .LP \fB2\fR \fBCall set\(hyup delay\fR .sp 1P .RT .PP Call set\(hyup delay applies only to the virtual call capability of packet\(hyswitched networks. .PP Call set\(hyup delay observed at a single section boundary, \fIB\fR\d\fIi\fR\u, is defined first and then call set\(hyup delay between a pair of section boundaries (\fIB\fR\d\fIi\fR\u, \fIB\fR\d\fIj\fR\u) is defined based on the former definition. In the former case, the call set\(hyup delay includes the delay for all virtual connection sections on the called user side of \fIB\fR\d\fIi\fR\uand the called user response time. In the latter case, the call set\(hyup delay includes only the delays between \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\u. Values are specified for call set\(hyup delay observed between section boundaries. .bp .RT .sp 1P .LP 2.1 \fIDefinition of call set\(hyup delay at a single section boundary\fR .sp 9p .RT .PP Call set\(hyup delay at a section boundary, \fIB\fR\d\fIi\fR\u, is defined using two Recommendation\ X.134 packet layer reference events (PEs). It is the period of time that starts when either a call request or an incoming call packet creates a PE at \fIB\fR\d\fIi\fR\u, and ends when the corresponding call connected or call accepted packet, accepting the virtual call, returns and creates its PE at \fIB\fR\d\fIi\fR\u. .PP Call set\(hyup delay at a section boundary = {\fIt\fR\d2\u\(em \fIt\fR\d1\u} where .RT .LP \fIt\fR\d1\u= Time of occurrence for the first PE. .LP \fIt\fR\d2\u= Time of occurrence for the second PE. .PP The two PEs can occur at any single section boundary within a virtual connection. The identities of the packets depend on the boundary of interest, as shown in Figure\ 2/X.135. The first packet is the call request packet and the second packet is the corresponding call connected packet at every boundary except the two boundaries that delimit the access circuit section associated with the called DTE. The first packet is the incoming call packet and the second packet is the call accepted packet at the latter two boundaries. The specific X.134 PEs used in measuring call set\(hyup delay at each section boundary are identified in Table\ 2/X.135. .LP .rs .sp 34P .ad r \fBFigure 2/X.135, (N), p.3\fR .sp 1P .RT .ad b .RT .LP .bp .ce \fBH.T. [T2.135]\fR .ce TABLE\ 2/X.135 .ce \fBPacket layer reference events (PEs) used\fR .ce \fBin measuring call set\(hyup delay\fR .ce \|\ua\d\u)\d .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(90p) | cw(36p) | cw(36p) . T{ X.134 packet layer reference event Circuit section T} Starting PE Ending PE _ .T& lw(90p) | cw(36p) | cw(36p) . T{ Calling DTE access circuit section T} 2 (X.25) 3 (X.25) _ .T& lw(90p) | cw(36p) | cw(36p) . T{ Called DTE access circuit section T} 1 (X.25) 4 (X.25) _ .T& lw(90p) | cw(36p) | cw(36p) . Internetwork circuit section 1 (X.75) T{ 2 (X.75) \ua\d\u)\d The PE numbers in this table refer to Tables 1/X.134 and 2/X.134 in Recommendation\ X.134. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 2/X.135 [T2.135], p.4\fR .sp 1P .RT .ad b .RT .sp 1P .LP 2.2 \fIDefinition of\fR \fBcall set\(hyup delay between two section boundaries\fR .sp 9p .RT .PP For a particular virtual call, call set\(hyup delay can be measured at one boundary, \fIB\fR\d\fIi\fR\u, and measured at another boundary, \fIB\fR\d\fIj\fR\u, further from the calling DTE. The difference in the values obtained is the call set\(hyup delay contributed by the virtual connection section(s) between the two boundaries. .PP Call set\(hyup delay between two section boundaries = {\fId\fR\d1\u\ \(em\ \fId\fR\d2\u} where .RT .LP \fId\fR\d1\u= Call set\(hyup delay measured at \fIB\fR\d\fIi\fR\u. .LP \fId\fR\d2\u= Call set\(hyup delay measured at \fIB\fR\d\fIj\fR\u. .PP The \fBend\(hyto\(hyend call set\(hyup delay\fR is the call set\(hyup delay between DTE boundaries, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d\fIn\fR\uin Figure\ 2/X.135. This end\(hyto\(hyend delay excludes the called user response time. The \fBnational\fR \fBportion call set\(hyup delay\fR is the call set\(hyup delay between the boundaries delimiting a national portion, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d5\uin Figure\ 2/X.135. The \fBinternational portion call set\(hyup delay\fR is the call set\(hyup delay between the boundaries delimiting an international portion, e.g.,\ \fIB\fR\d5\uand \fIB\fR\d\fIn\fR\\d\\u(em\d2\uin Figure\ 2/X.135. .sp 1P .LP 2.3 \fIValues\fR .sp 9p .RT .PP Table 3/X.135 defines worst\(hycase call set\(hyup delay values for each of the four virtual connection portion types identified in Table\ 1/X.135. DTE to DTE call set\(hyup delay values for two hypothetical reference connections are calculated in Annex\ C. All values are based on (and only apply under) the following assumptions .FS Values for other conditions are for further study. In the case of extremely long access lines and/or excessive delays in the access circuit section transmission equipment, these values may be exceeded. .FE : .RT .LP 1) Normal busy hour load conditions for the observed virtual connection. The definition of \*Qnormal busy hour load\*U as a traffic description is for further study. .LP 2) A basic call, in which none of the optional user facilities defined in Recommendation\ X.25 are used and no call user data is sent. .LP 3) Data link layer windows of entities outside the portion being specified are open (not flow controlled). .bp .PP The defined values consist of mean and 95% probability values. The mean is the expected value of the call set\(hyup delay distribution. The 95% probability value is the value below which 95% of the call set\(hyup delay values lie. Call set\(hyup attempts that are unsuccessful under the conditions of Recommendation\ X.136 are excluded and are addressed separately in that Recommendation. .ce \fBH.T. [T3.135]\fR .ce TABLE\ 3/X.135 .ce \fBWorst\(hycase call set\(hyup delay values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . mean (ms) 1000 + \fIX\fR 1600 + \fIX\fR 250 1600 _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . 95% (ms) 1500 + \fIX\fR 2100 + \fIX\fR 250 1800 _ .TE .nr PS 9 .RT .ad r \fBTableau 3/X.135 [T3.135], p.\fR .sp 1P .RT .ad b .RT .PP In Table 3/X.135, the value \fIX\fR \| depends on the signalling rate of the access circuit section that is included in the national portion. Table\ 4/X.135 presents the \fIX\fR values for user classes of service\ 8\(hy11 in Recommendation\ X.1 .FS These \fIX\fR \ values are not intended to represent the delay performance of the access circuit section, since these values do not include propagation delays, multiplexing delays, or the effects of retransmission. .FE . The \fIX\fR \ values for other signalling rates may be computed using the formula \v'6p' .sp 1P .ce 1000 \fIX\fR = 400/\fIR\fR ms, .ce 0 .sp 1P .LP .sp 1 where \fIR\fR \| is the signalling rate in kilobits per second. .FS The formula assumes that the transfer of each call set\(hyup packet across an access circuit section involves the transmission of 25\ octets: 5\ octets of frame level overhead, a 5\(hyoctet packet header, and 15\ octets of DTE address information .FE . .ce \fBH.T. [T4.135]\fR .ce TABLE\ 4/X.135 .ce \fBX\(hyvalues for Table 3/X.135\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(36p) | cw(36p) | cw(48p) . X.1 user class of service \fIR\fR (kbit/s) \fIX\fR (Milliseconds) _ .T& cw(36p) | cw(36p) | cw(48p) . \ 8 \ 2.4 167 _ .T& cw(36p) | cw(36p) | cw(48p) . \ 9 \ 4.8 \ 84 _ .T& cw(36p) | cw(36p) | cw(48p) . 10 \ 9.6 \ 42 _ .T& cw(36p) | cw(36p) | cw(48p) . 11 48.0 \ \ 9 _ .TE .nr PS 9 .RT .ad r \fBTableau 4/X.135 [T4.135], p.\fR .sp 1P .RT .ad b .RT .LP .bp .PP The call set\(hyup delay values defined in Table 3/X.135 are intended to be used as worst\(hycase limits in planning international packet\(hyswitched services. The actual delay performance achieved on a virtual connection portion will depend on many factors, including the traffic expected and actually offered, the internal network topology, and the signalling rates on the internetwork circuit sections. Variation away from the worst\(hycase value for each factor can improve the performance. .PP The overall call set\(hyup delay value for a set of concatenated virtual connection portions can be calculated directly by adding the individual portion means defined in Table\ 3/X.135. A method of calculating an overall 95% probability call set\(hyup delay value for a set of concatenated virtual connection portions from the individual 95% probability values is described in Annex\ C. .RT .sp 2P .LP \fB3\fR \fBData packet transfer delay\fR .sp 1P .RT .PP This delay refers to successful transfer of data packets and applies to both the virtual call and the permanent virtual circuit capabilities of packet\(hyswitched networks. It is defined only between pairs of section boundaries. .RT .sp 1P .LP 3.1 \fBdata packet transfer delay\fR \fIdefinition\fR .sp 9p .RT .PP Data packet transfer delay is the period of time that starts when a data packet creates a PE at a particular boundary, \fIB\fR\d\fIi\fR\u, and ends when this same packet creates a later PE at another boundary, \fIB\fR\d\fIj\fR\u. The specific X.134\ PEs used in measuring data packet transfer delay at each section boundary are identified in Table\ 5/X.135. .PP Data packet transfer delay = {\fIt\fR\d2\u\(em \fIt\fR\d1\u} where .RT .LP \fIt\fR\d1\u= Time of occurrence for the first PE. .LP \fIt\fR\d2\u= Time of occurrence for the second PE. .ce \fBH.T. [T5.135]\fR .ce TABLE\ 5/X.135 .ce \fBPacket layer reference events (PEs) used\fR .ce \fBin measuring data packet transfer delay\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(84p) | cw(48p) . T{ X.134 packet layer reference event Circuit section T} Starting/Ending PE _ .T& lw(84p) | cw(48p) . Source access circuit section 10a (X.25) _ .T& lw(84p) | cw(48p) . T{ Destination access circuit section T} 9a (X.25) _ .T& lw(84p) | cw(48p) . Internetwork circuit section 5a (X.75) _ .TE .nr PS 9 .RT .ad r \fBTableau 5X/.135 [T5.135], p.\fR .sp 1P .RT .ad b .RT .LP .sp 2 .PP The \fBend\(hyto\(hyend data packet transfer delay\fR is the one\(hyway delay between DTE boundaries, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d\fIn\fR\uin Figure\ 2/X.135. The \fBnational portion data packet transfer delay\fR is the delay between the boundaries delimiting a national portion, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d5\uin Figure\ 2/X.135. The \fBinternational portion data packet transfer delay\fR is the delay between the boundaries delimiting an international portion, e.g.,\ \fIB\fR\d5\uand \fIB\fR\d\fIn\fR\\d\\u(em\d2\uin Figure\ 2/X.135. .bp .sp 1P .LP 3.2 \fIValues\fR .sp 9p .RT .PP Table 6/X.135 defines worst case data packet transfer delay values for each of the four virtual connection portion types identified in Table\ 1/X.135. DTE to DTE data packet transfer delay values for two hypothetical reference connections are calculated in Annex\ C. All values are based on (and only apply under) the following assumptions .FS Values for other conditions are for further study. In the case of extremely long access lines and/or excessive delays in the access circuit section transmission equipment, these values may be exceeded. .FE : .RT .LP 1) Normal busy hour load conditions for the observed virtual connection. The definition of \*Qnormal busy hour load\*U as a traffic description is for further study. .LP 2) A user data field length of 128 octets. .LP 3) Data link and packet layer windows on the receiving DTE side of the portion being specified are open. .PP The defined values consist of mean and 95% probability values. The mean is the expected value of the data packet transfer delay distribution, excluding values that exceed a specified maximum data packet transfer delay. The 95% probability value is the value below which 95% of the data packet transfer delay values lie. Data packet transfer attempts that are unsuccessful under the conditions of Recommendation\ X.136 are excluded and are addressed separately in that Recommendation. .ce \fBH.T. [T6.135]\fR .ce TABLE\ 6/X.135 .ce \fBWorst\(hycase data packet transfer delay values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . mean (ms) 350 + \fIY\fR 650 + \fIY\fR 215 \ 950 _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . 95% (ms) 525 + \fIY\fR 825 + \fIY\fR 215 1125 _ .TE .nr PS 9 .RT .ad r \fBTableau 6/X.135 [T6.135], p.\fR .sp 1P .RT .ad b .RT .LP .sp 2 .PP In Table 6/X.135, the value \fIY\fR \| depends on the signalling rate of the access circuit section that is included in the national portion. Table\ 7/X.135 presents the \fIY\fR \ values for user classes of service\ 8\(hy11 in Recommendation\ X.1 .FS These \fIY\fR \ values are not intended to represent the delay performance of the access circuit section, since these values do not include propagation delays, multiplexing delays, or the effects of retransmission. .FE . The \fIY\fR \ values for other signalling rates may be computed using the formula \v'6p' .sp 1P .ce 1000 \fIY\fR = 1088/\fIR\fR ms, .ce 0 .sp 1P .LP .sp 1 where\| \fIR\fR \| is the signalling rate in kilobits per second .FS The formula assumes that the transfer of a data packet across an access circuit section involves the transmission of 136\ octets: 5\ octets of frame level overhead, a 3\(hyoctet packet header, and 128\ octets of user data. .FE . .LP .sp 1 .bp .ce \fBH.T. [T7.135]\fR .ce TABLE\ 7/X.135 .ce \fBY\(hyvalues for Table 6/X.135\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(36p) | cw(36p) | cw(48p) . X.1 user class of service \fIR\fR (kbit/s) \fIY\fR (Milliseconds) _ .T& cw(36p) | cw(36p) | cw(48p) . \ 8 \ 2.4 453 _ .T& cw(36p) | cw(36p) | cw(48p) . \ 9 \ 4.8 227 _ .T& cw(36p) | cw(36p) | cw(48p) . 10 \ 9.6 113 _ .T& cw(36p) | cw(36p) | cw(48p) . 11 48.0 \ 23 _ .TE .nr PS 9 .RT .ad r \fBTableau 7/X.135 [T7.135], p.9\fR .sp 1P .RT .ad b .RT .PP The data packet transfer delay values defined in Table\ 6/X.135 are intended to be used as worst\(hycase limits in planning international packet\(hyswitched services. The actual delay performance achieved on a virtual connection portion will depend on many factors, including the traffic expected and actually offered, the internal network topology, and the signalling rates on the internetwork circuit sections. Variation away from the worst\(hycase value for each factor can improve the performance. .PP The overall mean data packet transfer delay value for a set of concatenated virtual connection portions can be calculated directly by adding the individual portion means defined in Table\ 6/X.135. A method of calculating an overall 95% probability data packet transfer delay value for a set of concatenated virtual connection portions from the individual 95% probability values is described in Annex\ C. .RT .sp 2P .LP \fB4\fR \fBThroughput parameters\fR .sp 1P .RT .PP This section defines three throughput parameters: throughput, steady\(hystate throughput, and throughput capacity. Values are specified for throughput capacity. .RT .sp 1P .LP 4.1 \fBthroughput\fR \fIdefinition\fR .sp 9p .RT .PP Throughput for a virtual connection section is the number of user data bits successfully transferred in one direction across that section per unit time .FS User data bits are the bits of the user data field in data packets of the X.25 or X.75 packet level (protocols and data above the packet level). Framing, routing, bit stuffing, error control, and other protocol fields introduced by all protocols at or below the packet level are excluded. .FE . Successful transfer means that no user data bits are lost, added, or inverted in transfer. .PP Assume: .RT .LP 1) That data packet \fIA\fR\d0\uis the final packet of a complete packet sequence crossing input boundary\ \fIB\fR\d\fIi\fR\u. .LP 2) That subsequently, \fIk\fR sequential data packets (\fIA\fR\d1\u, \fIA\fR\d2\u, .\|.\|.\ \fIA\fR\d\fIk\fR\u) forming the next complete packet sequence cross the input boundary\ \fIB\fR\d\fIi\fR\uimmediately following \fIA\fR\d0\u. .LP 3) That data packet \fI\*\|A\fR\d0\uis the final packet of the first complete packet sequence when it crosses output boundary\ \fIB\fR\d\fIj\fR\u. .LP 4) That packets \fI\*\|A\fR\d1\u, \fI\*\|A\fR\d2\u, .\|.\|.\ \fI\*\|A\fI\d\fIm\fR\u\| comprise the second complete packet sequence when it crosses output boundary\ \fIB\fR\d\fIj\fR\u. .PP The X.134 PEs used in measuring throughput are the same as those used in measuring data packet transfer delay, as identified in Table\ 5/X.135. .bp .PP Let: .RT .LP \fIt\fR\d1\u =\ Time of occurrence for the PE created by \fIA\fR\d0\uat \fIB\fR\d\fIi\fR\u. .LP \fIt\fR\d2\u =\ Time of occurrence for the PE created by \fIA\fR\d\fIk\fR\u\| at \fIB\fR\d\fIi\fR\u. .LP \fIt\fR\d3\u =\ Time of occurrence for the PE created by \fI\*\|A\fR\d0\uat \fIB\fR\d\fIj\fR\u. .LP \fIt\fR\d4\u =\ Time of occurrence for the PE created by \fI\*\|A\fI\d\fIm\fR\u\| at \fIB\fR\d\fIj\fR\u. .LP \fIf\fR (\fIA\fR\d\fIr\fR\u) =\ Number of user data bits in packet \fIA\fR\d\fIr\fR\u. .PP Then a throughput measurement of size \fIk\fR \| is defined as follows: \v'6p' .sp 1P .ce 1000 Throughput measurement = $$4o @ pile {\fIk\fR above sum above \fIr\fR =1 } @ \fIf\fR (\fIA\fR\d\fIr\fR\u) $$1uMAX [(\fIt\fR\d2\u\(em \fIt\fR\d1\u), (\fIt\fR\d4\u\(em \fIt\fR\d3\u)] $$1e .ce 0 .sp 1P .LP .sp 1 .PP Recommendation X.136 defines conditions under which a transfer of consecutive data packets is considered to be unsuccessful. Only successful throughput measurements should be included in the assessment of throughput performance. .sp 1P .LP 4.2 \fBsteady\(hystate throughput\fR \fIdefinition\fR .sp 9p .RT .PP The steady\(hystate throughput for a virtual connection is the value to which a throughput measurement converges as the duration of the observation period increases with statistically constant load on the virtual connection. Assuming successful transfer, steady\(hystate throughput is the same when measured at every pair of section boundaries of the virtual connection. Thus, assuming no user data bits are lost, added, or inverted in transfer, a steady\(hystate throughput measurement can be made at any single section boundary within a virtual connection: \v'6p' .RT .sp 1P .ce 1000 Steady\(hystate throughput measurement = $$4o @ pile {\fIk\fR above sum above \fIr\fR =1 } @ \fIf\fR (\fIA\fR\d\fIr\fR\u) $$1u(\fIt\fR\d2\u\(em \fIt\fR\d1\u) $$1e .ce 0 .sp 1P .LP .sp 1 where \fIt\fR\d1\u, \fIt\fR\d2\uand \fIf\fR (\fIA\fR\d\fIr\fR\u) are defined above .FS Ancillary information on steady\(hystate throughput measurement is provided in Annex\ B. .FE . .PP Alternatively, the above equation can be used to calculate steady\(hystate throughput with different definitions for \fIt\fR\d1\uand \fIt\fR\d2\u. Times\ \fIt\fR\d1\uand \fIt\fR\d2\ucan be chosen in advance of the measurement. In this case, let (\fIA\fR\d1\u, \fIA\fR\d2\u, .\|.\|.\ \fIA\fR\d\fIk\fR\u) be the set of all virtual connection data packets crossing boundary\ \fIB\fR (creating PEs in one direction) at or following time\ \fIt\fR\d1\ubut before time\ \fIt\fR\d2\u. Then the above equation still measures steady\(hystate throughput. .sp 1P .LP 4.3 \fBthroughput capacity\fR \fIdefinition\fR .sp 9p .RT .PP Let \fIB\fR\d\fIi\fR\u\| and \fIB\fR\d\fIj\fR\u\| be two virtual connection section boundaries. Assume steady\(hystate throughput is to be estimated with data packets flowing from \fIB\fR\d\fIi\fR\uto \fIB\fR\d\fIj\fR\u. Assume there is a statistically constant load, \fIL\fR , on the virtual connection section between \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\u. Then the throughput capacity of that section under load\ \fIL\fR is defined as the steady\(hystate throughput maximized over all offered combinations of virtual connection parameter settings and choices for the performance and loading outside\ \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\u. Measurement of throughput capacity for a section between boundaries\ \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\uis accomplished in the same way as measurement of steady\(hystate throughput. However, measurement of throughput capacity requires that the components outside of \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\uhave significantly higher throughput capacity under their respective loads than the throughput capacity being measured. .PP For the given statistically constant load \fIL\fR between \fIB\fR\d\fIi\fR\u\| and \fIB\fR\d\fIj\fR\u, and for a given set of testing arrangements, any measured steady\(hystate throughput is a lower bound for the throughput capacity. To improve the estimate, the experiment may be repeated with different testing arrangements outside of \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\u(see Annex\ B). .PP The end\(hyto\(hyend throughput capacity is the throughput capacity between DTE boundaries, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d\fIn\fR\uin Figure\ 2/X.135. The national portion throughput capacity is the throughput capacity between the boundaries delimiting a national portion, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d5\uin Figure\ 2/X.135. The international portion throughput capacity is the throughput capacity between the boundaries delimiting an international portion, e.g.,\ \fIB\fR\d5\uand \fIB\fR\d\fIn\fR\\d\\u(em\d2\uin Figure\ 2/X.135. .bp .RT .sp 1P .LP 4.4 \fIValues\fR .sp 9p .RT .PP Table 8/X.135 defines worst\(hycase throughput capacity values for each of the four virtual connection portion types identified in Table\ 1/X.135. DTE to DTE throughput capacity values for two hypothetical reference connections are calculated in Annex\ C. All values are based on (and only apply under) the following assumptions .FS Values for other conditions are for further study. .FE : .RT .LP 1) Normal busy hour load conditions for the observed virtual connection. The definition of \*Qnormal busy hour load\*U as a traffic description is for further study. No other traffic on the access circuit sections. .LP 2) 9600 bit/s signalling rates on the access circuit sections. Applicability of the specified throughput capacity values to lower access circuit section signalling rates is for further study. .LP 3) A user data field length of 128 octets. Requested throughput class corresponding to 9600\ bit/s. (Note that the throughput class finally applying to the call may be lower than the requested throughput class.) .LP 4) Packet layer window sizes of 2 and data link layer window sizes of 7 on the access circuit sections. .LP 5) \fID\fR \| bit not used (\fID\fR = 0). .LP 6) Values apply to either direction of transfer. .LP 7) No unavailability (as defined in Recommendation\ X.137) during the observation period. .LP 8) No resets or premature disconnects (as defined in Recommendation\ X.136) during the observation period. .LP 9) Throughput capacity sample sizes of 200 packets (in the case of the first measurement technique specified in \(sc\ 4.2) or 2\ minutes (in the case of the alternative measurement technique specified in\ \(sc\ 4.2). .ce \fBH.T. [T8.135]\fR .ce TABLE\ 8/X.135 .ce \fBWorst\(hycase throughout capacity values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . mean (bit/s) 3000 2400 2000 1800 _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . 95% (bit/s) 2400 2000 1800 1500 _ .TE .nr PS 9 .RT .ad r \fBTableau 8/X.135 [T8.135], p.\fR .sp 1P .RT .ad b .RT .PP The defined values consist of mean and 95% probability values. The mean is the expected value of the throughput capacity distribution. The 95% probability value is the value above which 95% of the throughput capacity measurements lie. .PP The throughput capacity values defined in Table 8/X.135 are intended to be used as worse\(hycase limits in planning international packet\(hyswitched services. The actual throughput capacity achieved in a virtual connection portion will depend on many factors, including the traffic expected and actually offered, the internal network topology, and the signalling rates on the internetwork circuit sections. Variation away from the worse\(hycase value for each factor can improve the performance. The throughput capacity values defined here will not necessarily be achieved concurrently with the delay values defined in Table\ 6/X.135. .bp .PP An upper bound for the throughput capacity of a set of concatenated virtual connection portions can be derived from the individual portion throughput capacities as follows. If a portion between boundaries\ \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIj\fR\uhas throughput capacity\ \fIT\fR\d1\uunder load\ \fIL\fR\d1\u, and a portion between boundaries\ \fIB\fR\d\fIk\fR\uand \fIB\fR\d\fIm\fR\uhas throughput capacity\ \fIT\fR\d2\uunder load\ \fIL\fR\d2\u, and those portions are concatenated so that \fIB\fR\d\fIj\fR\u\ =\ \fIB\fR\d\fIk\fR\uwith \fIL\fR\d1\uand \fIL\fR\d2\uunchanged, then the resulting portion has throughput capacity. \v'6p' .RT .sp 1P .ce 1000 \fIT\fR \(= MIN [\fIT\fR\d1\u, \fIT\fR\d2\u] .ce 0 .sp 1P .PP .sp 1 Further information on estimating the throughput capacity of a set of concatenated virtual connection portions is provided in Annex\ C. .sp 2P .LP \fB5\fR \fBClear indication delay\fR .sp 1P .RT .PP Clear indication delay applies only to the virtual call capability of packet\(hyswitched networks. It is defined only between a pair of section boundaries. .RT .sp 1P .LP 5.1 \fBclear indication delay\fR \fIdefinition\fR .sp 9p .RT .PP Clear indication delay is the period of time that starts when either a clear request packet or a clear indication packet creates a PE at a boundary, \fIB\fR\d\fIi\fR\u, and ends when the corresponding clear request or clear indication packet creates a later PE at another boundary, \fIB\fR\d\fIj\fR\u. The specific X.134 PEs used in measuring clear indication delay at each section boundary are identified in Table\ 9/X.135. .PP Clear indication delay = {\fIt\fR\d2\u\(em \fIt\fR\d1\u} where .RT .LP \fIt\fR\d1\u= Time of occurrence for the first PE. .LP \fIt\fR\d2\u= Time of occurrence for the second PE. .ce \fBH.T. [T9.135]\fR .ce TABLE\ 9/X.135 .ce \fBPacket layer reference events (PEs) used\fR .ce \fBin measuring clear indication delay\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(84p) | cw(48p) . T{ X.134 packet layer reference event Circuit section T} Starting/Ending PE _ .T& lw(84p) | cw(48p) . T{ Clearing DTE access circuit section T} 6 (X.25) _ .T& lw(84p) | cw(48p) . T{ Cleared DTE access circuit section T} 5 (X.25) _ .T& lw(84p) | cw(48p) . Internetwork circuit section 3 (X.75) _ .TE .nr PS 9 .RT .ad r \fBTableau 9/X.135 [T9.135], p.\fR .sp 1P .RT .ad b .RT .PP The \fBend\(hyto\(hyend clear indication delay\fR is the one\(hyway delay between DTE boundaries, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d\fIn\fR\uin Figure\ 2/X.135. The \fBnational portion clear indication delay\fR is the delay between the boundaries delimiting a national portion, e.g.,\ \fIB\fR\d1\uand \fIB\fR\d5\uin Figure\ 2/X.135. The \fBinternational portion clear indication delay\fR is the delay between the boundaries delimiting an international portion, e.g.,\ \fIB\fR\d5\uand \fIB\fR\d\fIn\fR\\d\\u(em\d2\uin Figure\ 2./X.135. .bp .sp 1P .LP 5.2 \fIValues\fR .sp 9p .RT .PP Table 10/X.135 defines worst case clear indication delay values for each of the four virtual connection portion types identified in Table\ 1/X.135. DTE to DTE clear indication delay values for two hypothetical reference connections are calculated in Annex\ C. All values are based on (and only apply under) the following assumptions .FS Values for other conditions are for further study. In the case of extremely long access lines and/or excessive delays in the access circuit section transmission equipment, these values may be exceeded. .FE : .RT .LP 1) Normal busy hour load conditions for the observed virtual connection. The definition of \*Qnormal busy hour load\*U as a traffic description is for further study. .LP 2) Data link layer windows on the cleared DTE side of the portion being specified are open. .LP 3) The extended format of the clear request packet is not used. .ce \fBH.T. [T10.135]\fR .ce TABLE\ 10/X.135 .ce \fBWorst\(hycase clear indication delay values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . mean (ms) 500 + \fIZ\fR \ 800 + \fIZ\fR 110 800 _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . 95% (ms) 750 + \fIZ\fR 1050 + \fIZ\fR 110 900 _ .TE .nr PS 9 .RT .ad r \fBTableau 10/X.135 [T10.135], p.\fR .sp 1P .RT .ad b .RT .PP The defined values consist of mean and 95% probability values. The mean is the expected value of the clear indication delay distribution, excluding values that exceed a specified maximum clear indication delay. The 95% probability value is the value below which 95% of the clear indication delay values lie. Unsuccessful call clear attempts are excluded and are addressed separately in Recommendation\ X.136. .PP In Table 10/X.135, the value \fIZ\fR \| depends on the signalling rate of the access circuit section that is included in the national portion. Table\ 11/X.135 presents the \fIZ\fR \ values for user classes of service\ 8\(hy11 in Recommendation\ X.1 .FS These \fIZ\fR \ values are not intended to represent the delay performance of the access circuit section, since these values do not include propagation delays, multiplexing delays, or the effects of retransmission. .FE . .PP The \fIZ\fR \| values for other signalling rates may be computed using the formula \v'6p' .RT .sp 1P .ce 1000 \fIZ\fR = 80/\fIR\fR ms .ce 0 .sp 1P .LP .sp 1 where \fIR\fR \| is the signalling rate in kilobits per second .FS The formula assumes that the transfer of each call clearing packet across an access circuit section involves the transmission of 10\ octets: 5\ octets of frame level overhead and 5\ octets of packet header information. .FE . .LP .bp .ce \fBH.T. [T11.135]\fR .ce TABLE\ 11/X.135 .ce \fBZ\(hyvalues for Table 10/X.135\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(36p) | cw(36p) | cw(48p) . X.1 user class of service \fIR\fR (kbit/s) \fIZ\fR (Milliseconds) _ .T& cw(36p) | cw(36p) | cw(48p) . \ 8 \ 2.4 34 _ .T& cw(36p) | cw(36p) | cw(48p) . \ 9 \ 4.8 17 _ .T& cw(36p) | cw(36p) | cw(48p) . 10 \ 9.6 \ 9 _ .T& cw(36p) | cw(36p) | cw(48p) . 11 48.0 \ 2 _ .TE .nr PS 9 .RT .ad r \fBTableau 11/X.135 [T11.135], p.13\fR .sp 1P .RT .ad b .RT .PP The clear indication delay values defined in Table 10/X.135 are intended to be used as worst\(hycase values in planning international packet\(hyswitched services. The actual delay performance achieved on a virtual connection portion will depend on many factors, including the traffic expected and actually offered, the internal network topology, and the signalling rates on the internetwork circuit sections. Variation away from the worst\(hycase value for each factor can improve the performance. .PP The overall mean clear indication delay value for a set of concatenated virtual connection portions can be calculated directly by adding the individual portion means defined in Table\ 10/X.135. A method of calculating an overall 95% probability clear indication delay value for a set of concatenated virtual connection portions from the individual 95% probability values is described in Annex\ C. \v'6p' .RT .ce 1000 ANNEX\ A .ce 0 .ce 1000 (to Recommendation X.135) .sp 9p .RT .ce 0 .ce 1000 \fBFactors to be specified in reporting\fR .sp 1P .RT .ce 0 .sp 1P .ce 1000 \fBthroughput performance\fR .ce 0 .sp 1P .PP Many factors affect the throughout capacity that can be obtained on a virtual connection section. .sp 1P .RT .sp 2P .LP A.1 \fISignalling rates\fR .sp 1P .RT .PP The choice of signalling rates on circuit sections bounds throughput. In general, faster signalling rates improve throughput. .RT .sp 1P .LP A.2 \fIInterface windows\fR .sp 9p .RT .PP The choice of window size has an effect on throughput. In general, larger window sizes improve throughput. For maximum throughput, each user\(hycontrollable window size should be optimized with respect to delays and retransmission rates. .RT .sp 1P .LP A.3 \fIPacket length\fR .sp 9p .RT .PP The choice of packet length has an effect on throughput. In general, the use of larger packets improves throughput. For maximum throughput, packet sizes should be optimized with respect to the known error properties of the access links. .bp .RT .sp 1P .LP A.4 \fIAdditional virtual connections\fR .sp 9p .RT .PP Throughput of a tested virtual connection is dependent on the number of additional virtual connections and the loading in each direction on each connection. Throughput per virtual connection decreases as the number of additional virtual connections or the loading on the individual connections increases. When stating the throughput capacity of a virtual connection portion, the number of additional active virtual connections on the access circuit sections should be specified. Also, the total throughput in each direction on those virtual connections should be reported. For example: .RT .LP \*QThe throughput capacity of a virtual connection on this international portion is at least 1.2\ kbit/s. There can be at most 4\ additional virtual connections transmitting in the same direction between the same two portion boundaries at the same throughput.\*U .sp 1P .LP A.5 \fITime\(hyof\(hyday\fR .sp 9p .RT .PP When measuring throughput it is assumed that the loads on many connection components cannot be user controlled or observed. However, it is assumed that those loads are correlated with time\(hyof\(hyday, day\(hyof\(hyweek, and holidays. Thus users can improve their throughput by transmitting at particular times. .RT .sp 1P .LP A.6 \fIDirection\fR .sp 9p .RT .PP If the direction of the measurement affects the throughput capacity, the direction should be specified when stating throughput capacity. Otherwise the capacities in the two directions will be assumed to be equal. .RT .sp 1P .LP A.7 \fIThroughput class\fR .sp 9p .RT .PP Network internal windows and acknowledgement schemes may or may not be a function of a virtual connection's requested or default throughput class. For maximum throughput and when measuring throughput capacity, the throughput class for the virtual connection should be set to the maximum allowed by the section being measured. Because the optimum throughput class is always the maximum allowable, a statement of throughput capacity need not explicitly specify the throughput class. .RT .sp 1P .LP A.8 \fID bit usage\fR .sp 9p .RT .PP If the \fID\fR \| bit is set to 1 during a throughput measurement, that fact should be reported. Otherwise, the \fID\fR \ bit setting need not be reported. .RT .sp 1P .LP A.9 \fIDelay\fR .sp 9p .RT .PP Throughput and data packet transfer delay are related. If the throughput is specified under a delay constraint, then the delay should be reported. .RT .sp 1P .LP A.10 \fIReporting throughput capacity\fR .sp 9p .RT .PP Throughput capacity reports should specify the values of the controllable factors that were in effect during the throughput capacity measurement. All factors listed in this Annex should be reported unless otherwise specified. A typical report might specify conditions as follows: .RT .LP \*QFor this connection the network throughput capacity is at least 4.1\ kbit/s. The capacity was measured using two 9.6\ kbit/s access circuit sections, data link layer window sizes of\ 7, packet layer window sizes of\ 2, and 128\ octet user data fields. No additional virtual connections were present on either of the access circuit sections. The capacity was measured during the busiest hour of the weekday. The average data packet transfer delay during the measurement period was 500\ milliseconds. The precision of the throughput measurement is plus or minus 0.1\ kbit/s.\*U .PP With such statements, the throughput capacity is more easily verified and more easily matched to the throughput needs of potential users. .bp .ce 1000 ANNEX\ B .ce 0 .ce 1000 (to Recommendation X.135) .sp 9p .RT .ce 0 .ce 1000 \fBAncillary information on\fR \fBthroughput measurement\fR .sp 1P .RT .ce 0 .sp 1P .ce 1000 \fBand the application of throughput capacity values\fR .ce 0 .sp 1P .PP The following points should be noted with regard to throughput measurement: .sp 1P .RT .LP \(em A measurement of steady\(hystate throughput requires a measurement size of \fIk\fR \ =\ 200\ packets. An alternative is to specify a value for the measurement time period (\fIt\fR\d2\u\ \(em\ \fIt\fR\d1\u) of 2\ minutes. .LP \(em When measuring steady\(hystate throughput, data packets \fIA\fR\d1\uthrough \fIA\fR\d\fIk\fR\u\| need not constitute a single complete packet sequence. .LP \(em One way of verifying successful transfer of the test sequence in a steady\(hystate throughput measurement is to transfer another complete packet sequence. .LP \(em Throughput\(hyrelated measurements should not be conducted with user data sequences with high density of binary \*Qones\*U to avoid biasing the results by the effects of bit stuffing. .PP The following describes one way of applying the throughput capacity parameter. The discussion uses throughput capacity to design an international circuit section. .PP Assuming: .RT .LP \fIm\fR = the mean throughput per call (for the duration of the call) .LP \fIn\fR = the total number of calls present at any time .LP \fIp\fR = the number of those calls requiring the throughput capacity at any instant in time .LP \fIb\fR = the bit rate of the international internetwork circuit section and .LP \fIT\fR = the throughput capacity objective per call .PP Then the bit rate \fIb\fR \| should be: \v'6p' .sp 1P .ce 1000 \fIb\fR \(>=" (\fIm\fR * \fIn\fR ) + \fIp\fR (\fIT\fR \(em \fIm\fR ) .ce 0 .sp 1P .PP .sp 1 The actual \fIm\fR , \fIn\fR , and \fIp\fR \| values may be network dependent and reflect basically the population of the access line speeds and their traffic characteristics. It is therefore recommended that the value of\ \fIb\fR is chosen considerably higher than the value of (\fIm\fR * \fIn\fR ). The number of logical channels assigned to international internetwork links should depend on the relationship of the values\ \fIb\fR and\ \fIm\fR . .ce 1000 ANNEX\ C .ce 0 .ce 1000 (to Recommendation X.135) .sp 9p .RT .ce 0 .sp 1P .ce 1000 \fBRepresentative\fR \fBend\(hyto\(hyend speed of service performance\fR .sp 1P .RT .ce 0 .sp 1P .PP This Annex provides two examples to illustrate how end\(hyto\(hyend (DTE to DTE) speed of service performance can be estimated from the individual virtual connection portion performance values specified in X.135. Two example concatenations of Type\ A and Type\ B virtual connection portions are defined. The end\(hyto\(hyend call set\(hyup delay, data packet transfer delay, throughput capacity, and clear indication delay are calculated for each example. Although alternative network models and statistical assumptions are possible, the methods presented in this Annex provide one practical way of estimating end\(hyto\(hyend performance from the performance of individual network portions. .sp 1P .RT .sp 1P .LP C.1 \fIDefinition of the example end\(hyto\(hyend connections\fR .sp 9p .RT .PP For ease of reference the two example end\(hyto\(hyend (i.e., DTE to DTE) connections presented in this Annex will be referred to as \*QType\ 1\*U and \*QType\ 2\*U configurations. These hypothetical, but representative, configurations use the portion boundaries and packet layer reference events described in X.134. Figure\ 2/X.135 shows the relevant network boundaries and Table\ 1/X.135 defines the virtual connection portion types. .bp .PP The Type 1 configuration is defined to be: .RT .ce \fBH.T. [T12.135]\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(24p) | lw(108p) | lw(24p) . .T& cw(24p) | lw(108p) | cw(24p) . DTE DTE .T& lw(42p) | lw(48p) | lw(42p) . .TE .nr PS 9 .RT .ad r \fBTableau [T12.135], p.\fR .sp 1P .RT .ad b .RT .PP The Type 2 configuration is defined to be: .ce \fBH.T. [T13.135]\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(24p) | lw(108p) | lw(24p) . .T& cw(24p) | lw(108p) | cw(24p) . DTE DTE .T& lw(42p) | lw(48p) | lw(42p) . .TE .nr PS 9 .RT .ad r \fBTableau [T13.135], p.\fR .sp 1P .RT .ad b .RT .sp 1P .LP C.2 \fIEnd\(hyto\(hyend speed of service performance for the Type 1 and Type 2\fR \fIconfiguration examples\fR .sp 9p .RT .PP End\(hyto\(hyend speed of service performance values have been calculated for the example Type\ 1 and Type\ 2 connection configurations and are reported below in Tables\ C\(hy1/X.135 and\ C\(hy2/X.135. These calculations have been made by applying the methods derived in \(sc\ C.3 (below) to the individual network portions that, for convenience in defining these examples, are characterized by the worst\(hycase speed of service performance values specified in X.135. .PP The end\(hyto\(hyend performance for the mean call set\(hyup delay, data packet transfer delay, and clear indication delay are computed by simply summing the mean delays associated with the appropriate individual network portions. .RT .LP \fIExample\fR \ \(em\ For the Type 1 configuration the end\(hyto\(hyend mean call set\(hyup delay in milliseconds is computed by referring to Table\ 3/X.135 and adding the mean values for the National\ A and International\ A portion types: \v'6p' .sp 1P .ce 1000 (1000\ +\ \fIX\fR )\ +\ (250)\ +\ (1000\ +\ \fIX\fR )\ = 2250\ + 2\ * \ \fIX\fR .ce 0 .sp 1P .LP .sp 1 .PP The end\(hyto\(hyend performance for the 95th percentile call set\(hyup delay, data packet transfer delay, and clear indication delay can be determined by assuming (see \(sc\ C.3) that the variance of the end\(hyto\(hyend delay is the sum of the variances of the individual network portion delays. .LP \fIExample\fR \ \(em\ For the Type 1 configuration, referring to Table\ 3/X.135 and \(sc\ C.3, the 95th\ percentile value for the end\(hyto\(hyend call set\(hyup delay in milliseconds is: \v'6p' .ce 1000 (2250\|+\|2\| * \|\fIX\fR )\|+ [((1500\|+\|\fIX\fR )\|\(em\|(1000\|+\|\fIX\fR ))\u2\d\|+\|((250)\|\(em\|(250))\u2\d\|+ ((1500\|+\|X\fR )\|\(em\|(1000\|+\|\fIX\fR ))\u2\d]\u0\d\u.\d\u5\d .ce 0 .sp 1P .ce 1000 =\ 2957\ +\ 2\ * \ \fIX\fR .ce 0 .sp 1P .PP .sp 1 The end\(hyto\(hyend performance for the mean and 95th percentile for throughput capacity are determined by assuming that: .LP 1) the end\(hyto\(hyend throughput at any particular time is the minimum taken over all the individual network portions; and .LP 2) the throughput of an individual network portion is an independent and normally distributed random variable. \(sc\ C.3 derives formulas that combine the overlapping individual probability distributions to give the end\(hyto\(hyend throughput capacity distribution. .LP \fIExample\fR \ \(em\ Numerical computations of the end\(hyto\(hyend mean and 95th percentile throughput capacities for the Type\ 1 and Type\ 2 configurations are provided as examples in \(sc\ C.3.2. .bp .ce \fBH.T. [T14.135]\fR .ce TABLE\ C\(hy1/X.135 .ce \fBEnd\(hyto\(hyend speed of service performance\fR .ce \fBfor the type 1 configuration example\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(72p) | cw(42p) sw(42p) , ^ | c | c. Statistic Type 1 configuration Mean 95%ile _ .T& lw(72p) | lw(42p) | lw(42p) . Call set\(hyup delay (ms) 2250 + 2 * \fIX\fR 2957 + 2 * \fIX\fR _ .T& lw(72p) | lw(42p) | lw(42p) . T{ Data packet transfer delay (ms) T} \ 915 + 2 * \fIY\fR 1162 + 2 * \fIY\fR _ .T& lw(72p) | lw(42p) | lw(42p) . Throughput capacity (bit/s) 1999 1800 _ .T& lw(72p) | lw(42p) | lw(42p) . Clear indication delay (ms) 1110 + 2 * \fIZ\fR 1464 + 2 * \fIZ\fR _ .TE .nr PS 9 .RT .ad r \fBTableau C\(hy1/X.135 [T14.135], p.16\fR .sp 1P .RT .ad b .RT .LP .sp 5 .ce \fBH.T. [T15.135]\fR .ce TABLE\ C\(hy2/X.135 .ce \fBEnd\(hyto\(hyend speed of service performance\fR .ce \fBfor the Type 2 configuration example\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(72p) | cw(42p) sw(42p) , ^ | c | c. Statistic Type 2 configuration Mean 95%ile _ .T& lw(72p) | lw(42p) | lw(42p) . Call set\(hyup delay (ms) 4200 + 2 * \fIX\fR 4935 + 2 * \fIX\fR _ .T& lw(72p) | lw(42p) | lw(42p) . T{ Data packet transfer delay (ms) T} 1950 + 2 * \fIY\fR 2284 + 2 * \fIY\fR _ .T& lw(72p) | lw(42p) | lw(42p) . Throughput capacity (bit/s) 1797 1500 _ .T& lw(72p) | lw(42p) | lw(42p) . Clear indication delay (ms) 2100 + 2 * \fIZ\fR 2467 + 2 * \fIZ\fR _ .TE .nr PS 9 .RT .ad r \fBTableau C\(hy1/X.135 [T15.135], p.17\fR .sp 1P .RT .ad b .RT .LP .rs .sp 03P .ad r BLANC .ad b .RT .LP .bp .PP The parameters \fIX\fR , \fIY\fR \| and \fIZ\fR \| depend on the signalling rate of the access circuit section that is included in the national portion. Definitions, relevant assumptions, and values for \fIX\fR , \fIY\fR , and \fIZ\fR can be found in the appropriate sections of X.135. As noted in \(sc\ 4.4 of X.135, a 9.6\ kbit/s signalling rate for the access circuit sections is assumed for the worst\(hycase throughput capacity performance values. .sp 1P .LP C.3 \fIMethods for calculating mean and 95% points of delays and\fR \fIthroughputs of packet\(hyswitched services with two or more concatenated\fR \fIportions\fR .sp 9p .RT .PP This section describes the methods used in calculating end\(hyto\(hyend speed of service performance from individual network portion performance values. .RT .sp 1P .LP C.3.1 \fIDelays\fR .sp 9p .RT .PP It is assumed that a packet\(hyswitched service has \fIn\fR \| portions with delays \fId\fR\d1\u, \fId\fR\d2\u, .\|.\|., \fId\fR\d\fIn\fR\uvarying randomly with means\ \fIm\fR\d1\u, \fIm\fR\d2\u, .\|.\|., \fIm\fR\d\fIn\fR\uand 95% points\ \fIz\fR\d1\u, \fIz\fR\d2\u, .\|.\|., \fIz\fR\d\fIn\fR\u. Then the total delay \fID\fR \ =\ \fId\fR\d1\u\ +\ \fId\fR\d2\u\ +\ \|.\|.\|.\ +\ \fId\fR\d\fIn\fR\uhas a distribution with mean \v'6p' .RT .sp 1P .ce 1000 \fIM\fR = \fIm\fR\d1\u+ \fIm\fR\d2\u+ .\|.\|. + \fIm\fR\d\fIn\fR\u .ce 0 .sp 1P .LP .sp 1 .LP (with no further assumption). In order to obtain the 95% point of \fID\fR \| it is assumed that the delays\ \fId\fR\d\fIi\fR\uare statistically independent and that \fIz\fR\d\fIi\fR\u\ =\ \fIm\fR\d\fIi\fR\u\ +\ \fIk\fR \(*s\fI\fI\d\fIi\fR\uwith the same \fIk\fR for all portions, where \(*s\fI\fI\d\fIi\fR\uis the standard deviation of \fId\fR\d\fIi\fR\u. The like equality is also assumed for \fID\fR , i.e.,\ \fIZ\fR \ =\ \fIM\fR \ +\ \fIk\fR \(*s\fI\fI\d\fID\fR\u, where \fIZ\fR is the 95% point of\ \fID\fR . These equalities are true for normal distributions with \fIk\fR \ =\ 1.645. Then the variance of\ \fID\fR is the sum of the variances of the \fId\fR\d\fIi\fR\u. It follows that the 95% point of\ \fID\fR is given by \v'6p' .sp 1P .ce 1000 \fIZ\fR = \fIM\fR + [(\fIz\fR\d1\u\(em \fIm\fR\d1\u)\u2\d + (\fIz\fR\d2\u\(em \fIm\fR\d2\u)\u2\d + .\|.\|. + (\fIz\fR\d\fIn\fR\u\(em \fIm\fR\d\fIn\fR\u)\u2\d]\u1\d\u/\d\u2\d .ce 0 .sp 1P .PP .sp 1 The assumption of normality seems reasonable, but other assumptions are possible and could give substantially different answers. .sp 1P .LP C.3.2 \fIThroughputs\fR .sp 9p .RT .PP It is assumed that a packet\(hyswitched service has \fIn\fR \| portions with throughputs\ \fIT\fR\d1\u, \fIT\fR\d2\u, .\|.\|., \fIT\fR\d\fIn\fR\u\| varying randomly and independently with means \fIM\fR\d1\u, \fIM\fR\d2\u, .\|.\|., \fIM\fR\d\fIn\fR\u\| and 5%\ points (points exceeded by 95% of the values) \fIZ\fR\d1\u, \fIZ\fR\d2\u, .\|.\|., \fIZ\fR\d\fIn\fR\u. The net throughput of the service is assumed to be \fIV\fR \ =\ min (\fIT\fR\d1\u, \fIT\fR\d2\u, .\|.\|., \fIT\fR\d\fIn\fR\u). The cumulative distribution function (cdf) of \fIT\fR\d\fIi\fR\uis the probability that \fIT\fR\d\fIi\fR\uis less than or equal to any value, say\ \fIt\fR , and is denoted by \fIF\fR\d\fIi\fR\u(\fIt\fR ): \v'6p' .RT .sp 1P .ce 1000 \fIF\fR\d\fIi\fR\u(\fIt\fR ) = Prob [\fIT\fR\d\fIi\fR\u\(= \fIt\fR ], \fIi\fR = 1, 2, .\|.\|., \fIn\fR .ce 0 .sp 1P .LP .sp 1 .PP The probability density function (pdf) of \fIT\fR\d\fIi\fR\u\| is the derivative of \fIF\fR\d\fIi\fR\u(\fIt\fR ) and is denoted by \fIf\fR\d\fIi\fR\u(\fIt\fR )\ =\ \fIdF\fI\d\fIi\fR\u/\fIdt\fR . .PP In order to calculate the mean, say \fIM\fR\d\fIV\fR\\d\fIn\fR\u, and the 5% point, \fIV\fR \d.05, \fIn\fR \u , of the net throughput \fIV\fR , it is in general not sufficient to consider just the portion\ \fIM\fR\d\fIi\fR\u's and \fIZ\fR\d\fIi\fR\u's; it is necessary to combine the entire distributions\ \fIF\fR\d\fIi\fR\u(\fIt\fR ) (or \fIf\fR\d\fIi\fR\u(\fIt\fR )) to obtain the pdf of \fIV\fR , to be denoted by \fIg\fR\d\fIn\fR\u(\fIv\fR ). However, in the important special case that the portion with the usually smallest throughput (the \*Qslowest portion\*U) has a distribution that is not overlapped at all by the distributions of the larger throughputs, then the net throughput distribution is identical with that of the slowest portion, having the same mean and 5%\ point in particular. If the overlap of any other distribution with the slowest portion's distribution is negligible, then the same conclusion can be drawn. Later examples will suggest how much overlap can be considered negligible. .PP The case of general distributions is now resumed, that with \fIn\fR \ =\ 2 at first. Integration in the two dimensions of (\fIT\fR\d1\u, \fIT\fR\d2\u) shows that the pdf of \fIV\fR is given by \v'6p' .RT .ce 1000 \fIg\fR\d2\u(\fIv\fR ) = \fIf\fR\d1\u(\fIv\fR ) [1 \(em \fIF\fR\d2\u(\fIv\fR )] + \fIf\fR\d2\u(\fIv\fR ) [1 \(em \fIF\fR\d1\u(\fIv\fR )] .ce 0 .ad r (C\(hy1) .ad b .RT .LP .sp 1 .bp .LP The mean net throughput of the service is then \v'6p' .ce 1000 \fIM\fR\d\fIV\fR\\d2\u= @ pile {\(if above int above 0 } @ \fIvg\fR\d2\u(\fIv\fR ) \fIdv\fR .ce 0 .ad r (C\(hy2) \v'1P' \v'2p' .ad b .RT .LP .sp 1 and the 5% point is the value \fIV\fR \d.05, 2 \u such that \v'6p' .ce 1000 @ pile {\fIV\fR \d.05, 2 \u above int above 0 } @ \fIg\fR\d2\u(\fIv\fR ) \fIdv\fR = 0.05 .ce 0 .ad r (C\(hy3) \v'1P' \v'2p' .ad b .RT .LP .sp 1 .LP If \fIf\fR\d1\u(\fIt\fR ) = \fIf\fR\d2\u(\fIt\fR ), then \v'6p' .ce 1000 \fIg\fR\d2\u(\fIv\fR ) = 2\fIf\fR\d1\u(\fIv\fR ) [1 \(em \fIF\fR\d1\u(\fIv\fR )] .ce 0 .ad r (C\(hy4) .ad b .RT .PP .sp 1 It is now assumed that the portion throughput distributions are normal and that they are sufficiently concentrated that the tail of the fitted normal distribution to the left to zero is negligible (as is true for all the numerical values in X.135). The assumption is expressed in terms of the standard normal pdf \(*f (\fIu\fR ) and cdf \(*u (\fIx\fR ): \v'6p' .ce 1000 \(*f (\fIu\fR ) = @ {1 } over { sqrt {2\(*p } } @ \fIe\fR \u\(em\fIu\fR 2 /2 \d, \(*u (\fIx\fR ) = @ pile {\fIx\fR above int above \(em\(if } @ \(*f (\fIu\fR ) \fIdu\fR .ce 0 .ad r (C\(hy5) \v'1P' \v'2p' .ad b .RT .LP .sp 1 .LP Then \v'6p' .ce 1000 \fIf\fR\d\fIi\fR\u(\fIt\fR ) = @ {1 } over {\(*s\fI\fI\d\fIi\fR\u} @ \(*f @ left ( {\fIt\fR \(em \fIM\fR\d\fIi\fR\u} over {\(*s\fI\fI\d\fIi\fR\u} right ) @ , \fIF\fR\d\fIi\fR\u(\fIt\fR ) = @ pile {\fIt\fR above int above \(em\(if } @ \fIf\fR\d\fIi\fR\u( \fIy\fR ) \fIdy\fR .ce 0 .ad r (C\(hy6) \v'1P' \v'2p' .ad b .RT .LP .sp 1 where the standard deviation \(*s\fI\fI\d\fIi\fR\u= (\fIM\fR\d\fIi\fR\u\(em \fIZ\fR\d\fIi\fR\u)/1.64485. In the case \fIf\fR\d1\u(\fIt\fR )\ =\ \fIf\fR\d2\u(\fIt\fR ), then \v'6p' .ce 1000 \fIg\fR\d2\u(\fIv\fR ) = @ {2 } over {\(*s\d1\u} @ \(*f @ left ( {\fIv\fR \(em \fIM\fR\d1\u} over {\(*s\d1\u} right ) @ @ left [1 \(em \(*u left ( {\fIv\fR \(em \fIM\fR\d1\u} over {\(*s\d1\u} right ) right ] @ .ce 0 .ad r (C\(hy7) \v'8p' .ad b .RT .LP .sp 1 .PP The case \fIn\fR \ =\ 3 is now considered. The pdf \fIg\fR\d3\u(\fIv\fR ) of \fIV\fR\d3\u\ =\ min (\fIT\fR\d1\u, \fIT\fR\d2\u, \fIT\fR\d3\u) can be obtained by iteration on the distribution of \fIV\fR\d2\u\ =\ min (\fIT\fR\d1\u, \fIT\fR\d2\u) since \fIV\fR\d3\u\ =\ min (\fIV\fR\d2\u, \fIT\fR\d3\u). Hence \v'6p' .ce 1000 \fIg\fR\d3\u(\fIv\fR ) = \fIg\fR\d2\u(\fIv\fR ) [1 \(em \fIF\fR\d3\u(\fIv\fR )] + \fIf\fR\d3\u(\fIv\fR ) [1 \(em \fIG\fR\d2\u(\fIv\fR )] .ce 0 .ad r (C\(hy8) .ad b .RT .LP .sp 1 where g\fR\d2\u(\fIv\fR ) is given by (C\(hy1) and \fIG\fR\d2\u(\fIv\fR ) is its indefinite integral, \v'6p' .ce 1000 \fIG\fR\d2\u(\fIv\fR ) = @ pile {\fIv\fR above int above 0 } @ \fIg\fR\d2\u(\fIx\fR ) \fIdx\fR .ce 0 .ad r (C\(hy9) \v'1P' \v'2p' .ad b .RT .LP .sp 1 .LP If all three pdf's \fIf\fR\d\fIi\fR\u(\fIt\fR ) are identical, the \fIg\fR\d3\u(\fIv\fR ) simplifies to \v'6p' .ce 1000 \fIg\fR\d3\u(\fIv\fR ) = 3\fIf\fR\d1\u(\fIv\fR ) [1 \(em \fIF\fR\d1\u(\fIv\fR )]\u2\d .ce 0 .ad r (C\(hy10) .ad b .RT .LP .sp 1 .bp .LP Normal as well as identical distributions are now assumed. Then, from (C\(hy5), (C\(hy6), and (C\(hy10), \v'6p' .LP \fIM\fR\d\fIV\fR\\d3\u = @ pile {\(if above int above 0 } @ \fIvg\fR\d3\u(\fIv\fR ) \fIdv\fR .LP = \fIM\fR\d1\u+ 3\(*s\d1\u @ pile {\(if above int above \(em\(if } @ \fIu\fR \(*f (\fIu\fR ) [1 \(em \(*u (\fIu\fR )]\u2\d \fIdu\fR .LP = \fIM\fR\d1\u\(em 3\(*s\d1\u @ pile {\(if above int above 0 } @ \fIu\fR \(*f (\fIu\fR ) [2 \(*u (\fIu\fR ) \(em 1] \fIdu\fR .ad r = \fIM\fR\d1\u\(em \(*s\d1\u\fIK\fR\d3\u, (C\(hy11) .ad b .RT .LP .sp 1 .LP where \fIK\fR\d3\u= 0.8463 by Teichroew (1956). Likewise \v'6p' .ce 1000 \fIV\fR \d.05, 3 \u = \fIM\fR\d1\u+ \(*s\d1\u\fIU\fR \d.05, 3 \u .ce 0 .ad r (C\(hy12) .ad b .RT .LP .sp 1 where \v'6p' .ce 1000 3 @ pile {\fIU\fR \d.05, 3 \u above int above \(em\(if } @ \(*f (\fIu\fR ) [1 \(em \(*u (\fIu\fR )]\u2\d \fIdu\fR = 0.05 .ce 0 .ad r (C\(hy13) \v'1P' \v'2p' .ad b .RT .LP .sp 1 By integration \v'6p' .ce 1000 \(*u (\(em\fIU\fR \d,05, 3 \u) = 1 \(em 0.095\u1\d\u/\d\u3\d = 0.016952 .ce 0 .ad r (C\(hy14) .ad b .RT .LP .sp 1 .LP Hence from any cumulative normal distribution table, \fIU\fR \d.05, 3 \u\ =\ 2.121. .LP \fIExample\ 1\fR \ \(em\ Calculate the mean and 95th percentile net throughputs assuming there are three identical and normal portion distributions with \fIM\fR\d1\u\ =\ \fIM\fR\d2\u\ =\ \fIM\fR\d3\u\ =\ 2000\ bit/s and \fI\fR \fIZ\fR\d1\u\ =\ \fIZ\fR\d2\u\ =\ \fIZ\fR\d3\u\ =\ 1800\ bit/s. Then \(*s\d1\u\ =\ \(*s\d2\u\ =\ \(*s\d3\u\ =\ 2000/1.645\ =\ 121.6\ bit/s. From\ (C\(hy11): \v'6p' .sp 1P .ce 1000 \fIM\fR\d\fIV\fR\\d3\u= 2000 \(em 121.6 \(mu 0.8463 = 1897 bit/s .ce 0 .sp 1P .LP .sp 1 From (C\(hy12) and (C\(hy14) \v'6p' .sp 1P .ce 1000 \fIV\fR \d.05, 3 \u = 2000 \(em 121.6 \(mu 2.121 = 1742 bit/s .ce 0 .sp 1P .LP .sp 1 .LP \fIExample\ 2\fR \ \(em\ Consider the Type 1 configuration. From Table\ 8/X.135, \fIM\fR\d1\u\ =\ \fIM\fR\d2\u\ =\ 3000\ bit/s, \fIM\fR\d3\u\ =\ 2000\ bit/s, \fIZ\fR\d1\u\ =\ \fIZ\fR\d2\u\ =\ 2400\ bit/s, \fIZ\fR\d3\u\ =\ 1800\ bit/s. With normal distributions there is slight but probably negligible overlap of the larger throughputs with the smallest throughput; the probability of either national throughput being less than or equal to the \fIupper\fR 5%\ point of the international throughput, 2200\ bit/s, is 0.014. Hence, at least approximately, \fIM\fR\d\fIV\fR\\d3\u\ =\ \fIM\fR\d3\u\ =\ 2000\ bit/s, \fIV\fR \d.05, 3 \u\ =\ \fIZ\fR\d3\u\ =\ 1800\ bit/s. .PP This can be checked by numerical integration. Since this will come up in other applications, general formulas are given here. When \fIf\fR\d1\u(\fIv\fR )\ =\ \fIf\fR\d2\u(\fIv\fR ), \fIG\fR\d2\u(\fIv\fR ) in\ (C\(hy9) becomes \v'6p' .sp 1P .ce 1000 \fIG\fR\d2\u(\fIv\fR ) = 2\fIF\fR\d1\u(\fIv\fR ) \(em [\fIF\fR\d1\u(\fIv\fR )]\u2\d .ce 0 .sp 1P .LP .sp 1 .LP When the distributions are also normal, it follows from (C\(hy8) and (C\(hy5) that \v'6p' .sp 1P .ce 1000 \fIg\fR\d3\u(\fIv\fR ) = @ left [1 \(em \(*u left ( {\fIv\fR \(em \fIm\fR\d1\u} over {\(*s\d1\u} right ) right ] @ @ left { {2 } over {\(*s\d1\u} \(*f left ( {\fIv\fR \(em \fIm\fR\d1\u} over {\(*s\d1\u} right ) left [1 \(em \(*u left ( {\fIv\fR \(em \fIm\fR\d3\u} over {\(*s\d3\u} right ) right ] + .ce 0 .sp 1P .ce 1000 + {1 } over {\(*s\d3\u} \(*f left ( {\fIv\fR \(em \fIm\fR\d3\u} over {\(*s\d3\u} right ) left [1 \(em \(*u left ( {\fIv\fR \(em \fIm\fR\d1\u} over {\(*s\d1\u} right ) right ] right } @ .ce 0 .ad r (C\(hy15) \v'7p' .ad b .RT .LP .sp 1 .bp .LP Hence the mean throughput for a three\(hyportion network with two portions identical is, with the change of variable\ \fIu\fR \ =\ (\fIv\fR \ \(em\ \fIm\fR\d1\u)/\(*s\d1\u, \v'6p' .ce 1000 \fIM\fR\d\fIV\fR\\d3\u= @ pile {\(if above int above \(em\(if } @ (\fIm\fR\d1\u+ \(*s\d1\u\fIu\fR )[1 \(em \(*u (\fIu\fR )] @ left {\fIZ\fR \(*f (\fIu\fR ) left [1 \(em \(*u left ( {\fIm\fR\d1\u\(em \fIm\fR\d3\u+ \(*s\d1\u\fIu\fR } over {\(*s\d3\u} right ) right ] + .ce 0 .sp 1P .ce 1000 (C\(hy16) + {\(*s\d1\u} over {\(*s\d3\u} \(*f left ( {\fIm\fR\d1\u\(em \fIm\fR\d3\u+ \(*s\d1\u\fIu\fR } over {\(*s\d3\u} right ) [1 \(em \(*u (\fIu\fR )] right } @ \fIdu\fR .ce 0 .sp 1P .LP .PP This can be integrated numerically using a pocket calculator and the National Bureau of Standards \fITables of Normal Probability Functions\fR . Since these tables give the integral of \(*f (\fIu\fR ) from \(em\fIx\fR to \fIx\fR , say \fIS\fR (\fIx\fR ), rather than \(*u (\fIx\fR ), the following substitution is made in (C\(hy16) (in three places): \v'6p' .ce 1000 1 \(em \(*u (\fIu\fR ) = @ left { pile { {[1 \(em \fIS\fR BOCAD15\fR (\fIu\fR )]/2 } above { [1 + \fIS\fR (|\fIu\fR |)]/2 } } \ \ pile { {if \fIu\fR >=" 0 } above { if \fIu\fR < 0 } } .ce 0 .ad r (C\(hy17) \v'6p' .ad b .RT .LP .sp 1 .PP In the above Example 2, (C\(hy16) becomes \v'6p' .sp 1P .ce 1000 {2\fIM\fR\d\fIV\fR\\d3\u} over {\(*s\d1\u} = pile {\(if above int above \(em\(if } (8.225 + \fIu\fR ) [1 \(+- \fIS\fR (|\fIu\fR |)] {\(*f (\fIu\fR )[1 \(+- \fIS\fR (|8.225 + 3\fIu\fR |)] .ce 0 .sp 1P .ce 1000 + 1.5 \(*f (8.225 + 3\fIu\fR )[1 \(+- \fIS\fR (|\fIu\fR |)]} \fIdu\fR .ce 0 .sp 1P .PP .sp 1 Numerical integration with \(*D\fIu\fR = 0.1 and the Trapezoidal Rule yields \fIM\fR\d\fIV\fR\\d3\u\ =\ 1999.09\ bit/s. With Simpson's Rule\ \fIM\fR\d\fIV\fR\\d3\u\ =\ 1999.11\ bit/s. Hence the slight overlap of the distributions of the two larger throughputs with the smaller throughput distribution reduces the mean net throughput by less than 1\ bit/s. The effect on the lower 5%\ point will be much less, so \fIV\fR \d.05, 3 \u\ =\ 1800\ bit/s. However, comparison with Example\ 1 shows that \fIcomplete\fR overlap of three portion distributions does reduce the throughput substantially below that of an individual portion. .LP \fIExample\ 3\fR \ \(em\ Consider the Type 2 configuration. From Table 8/X.135, \fIM\fR\d1\u\ =\ 3000, \fIM\fR\d2\u\ =\ 2400, \fIM\fR\d3\u\ =\ 1800, \fIZ\fR\d1\u\ =\ 2400, \fIZ\fR\d2\u\ =\ 2000, \fIZ\fR\d3\u\ =\ 1500 (all bit/s). Three non\(hyidentical portions result in an integral substantially messier than (16). It could be programmed on a computer, but that is unnecessary because a tight bound can be obtained by replacing the fastest portion by one identical with the next faster portion and using (16). Doing so with \(*D\fIu\fR \ =\ 0.1 and the Trapezoidal Rule gives \fIM\fR\d\fIV\fR\\d3\u\ =\ 1794.4\ bit/s; the more accurate Simpson's Rule gives \fIM\fR\d\fIV\fR\\d3\u\ =\ 1794.7\ bit/s. Since \fIM\fR\d\fIV\fR\\d3\umust be less than or equal to \fIM\fR\d3\u\ =\ 1800\ bit/s, the mean throughput with the original three non\(hyidentical portions is bounded by 1795 and 1800\ bit/s. It is estimated as 1797\ bit/s with an error probably no more than\ 1\ bit/s. The effect on the lower 5%\ point will be even less; numerical integration with \(*D\fIu\fR \ =\ 0.1 gives \fIV\fR \d.05, 3 \u\ =\ 1499.2\ bit/s when the fastest portion is replaced by one identical with the next faster portion, so it is estimated that the original network has \fIV\fR \d.05, 3 \u\ =\ 1500\ bit/s to the nearest unit. .PP These examples suggest the following when the smallest throughput distribution is not greatly overlapped by others, and this applies no matter how many portions there are: .PP \fIGeneral Rule\fR \ \(em\ If the mean throughput of the slowest portion is less than the mean of the next slowest portion by at least twice the difference between the mean and 95%ile of the slowest portion or of the next slowest portion, whichever difference is larger, then the mean and 95%ile of the throughput of the network are the same as those of the slowest portion (with negligible error). (This rule can probably be relaxed by replacing \*Qtwice\*U by \*Q1.5\ times\*U or deleting \*Qtwice\*U without incurring too much error in practice.) .bp .PP The case of general \fIn\fR \| is considered similarly. With different distributions \fIf\fR\d\fIi\fR\u(\fIt\fR ) the pdf \fIg\fR\d\fIn\fR\u(\fIv\fR ) of \fIV\fR\d\fIn\fR\u\ =\ min (\fIT\fR\d1\u, \fIT\fR\d2\u, .\|.\|., \fIT\fR\d\fIn\fR\u) is obtainable by iteration from \fIg\fR\d\fIn\fR\\d\\u(em\d1\u\fR (\fIv\fR ): \v'6p' .RT .sp 1P .ce 1000 \fIg\fR\d\fIn\fR\u(\fIv\fR ) = \fIg\fR\d\fIn\fR\\d\\u(em\d1\u(\fIv\fR ) [1 \(em \fIF\fR\d\fIn\fR\u(\fIv\fR )] + \fIf\fR\d\fIn\fR\u(\fIv\fR ) [1 \(em \fIG\fR\d\fIn\fR\\d\\u(em\d1\u(\fIv\fR )] .ce 0 .sp 1P .LP .sp 1 If all \fIf\fR\d\fIi\fR\u(\fIt\fR ) are identical, then \v'6p' .sp 1P .ce 1000 \fIg\fR\d\fIn\fR\u(\fIv\fR ) = \fInf\fI\d\fIi\fR\u(\fIv\fR ) [1 \(em \fIF\fR\d\fIi\fR\u(\fIv\fR )]\fI\fI \u\fIn\fR\d\uD\dlF261\u1\d .ce 0 .sp 1P .LP .sp 1 .LP If, in addition, normal distributions are assumed for the \fIf\fR\d\fIi\fR\u(\fIt\fR ), then the mean net throughput is \v'6p' .LP \fIM\fR\d\fIV\fR\\d\fIn\fR\u = \fIM\fR\d1\u+ \fIn\fR \(*s\d1\upile {\(if above int above \(em\(if } \fIu\fR \(*f (\fIu\fR ) [1 \(em \(*u (\fIu\fR )] \u\fIn\fR \(em1 \d \fIdu\fR , .LP = \fIM\fR\d1\u\(em \fIn\fR \(*s\d1\upile {\(if above int above 0 } \fIu\fR \(*f (\fIu\fR ) {\(*u \u\fIn\fR \(em1 \d (\fIu\fR ) \(em [1 \(em \(*u (\fIu\fR )] \u\fIn\fR \(em1 \d } \fIdu\fR .ad r = \fIM\fR\d1\u\(em \fIK\fR\d\fIn\fR\u\(*s\d1\u (C\(hy18) .ad b .RT .LP and the 5% point of the net throughput is \v'6p' .ce 1000 \fIV\fR \d.05, \fIn\fR \u = \fIM\fR\d1\u\(em \(*s\d1\u\fIU\fR \d.05, \fIn\fR \u .ce 0 .ad r (C\(hy19) .ad b .RT .LP .sp 1 .LP where \v'6p' .ce 1000 \(*u (\(em\fIV\fR \d.05, \fIn\fR \u) = 1 \(em 0.95\u1\d\u/\d\fI\fI \u\fIn\fR\d .ce 0 .ad r (C\(hy20) .ad b .RT .LP .sp 1 The values \fIK\fR\d\fIn\fR\u\| and \fIU\fR \d.05, \fIn\fR \u\| can be tabulated as a function of \fIn\fR : .ce \fBH.T. [T16.135]\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(18p) | lw(18p) | lw(18p) | lw(18p) | lw(18p) | lw(18p) . .T& cw(18p) | cw(18p) | cw(18p) | cw(18p) | cw(18p) | cw(18p) . \fIn\fR 1 2 3 4 5 _ .T& lw(18p) | lw(18p) | lw(18p) | lw(18p) | lw(18p) | lw(18p) . \fIK\fI 0 0.5642 0.8463 1.0294 1.1630 _ .T& lw(18p) | lw(18p) | lw(18p) | lw(18p) | lw(18p) | lw(18p) . T{ \fIU\fR \d.05,\|\fIn\fR \u T} 1.645 1.955\ 2.121\ 2.234\ 2.319\ _ .TE .nr PS 9 .RT .ad r \fBTableau [T16.135], p.\fR .sp 1P .RT .ad b .RT .LP .sp 2 .sp 1P .LP C.4 \fINotes on key assumptions, results, and implications\fR .sp 9p .RT .PP For further study. .RT .sp 2P .LP \fBReference\fR .sp 1P .RT .LP [1] TEICHROEW, D., Tables of expected values of order statistics and products of order statistics for samples of size twenty and less from the normal distribution, \fIAnnals of Mathematical Statistics\fR , \fB27\fR , pp. 410\(hy426, 1956. .bp .sp 2P .LP \fBRecommendation\ X.136\fR .RT .sp 2P .ce 1000 \fBACCURACY\ AND\ DEPENDABILITY\ PERFORMANCE\ VALUES\| FOR\ PUBLIC\ DATA\ NETWORKS\fR .EF '% Fascicle\ VIII.3\ \(em\ Rec.\ X.136'' .OF '''Fascicle\ VIII.3\ \(em\ Rec.\ X.136 %' .ce 0 .sp 1P .ce 1000 \fBWHEN\ PROVIDING\ INTERNATIONAL\ PACKET\(hySWITCHED\ SERVICES\fR .ce 0 .sp 1P .ce 1000 \fI(Malaga\(hyTorremolinos, 1984; amended at Melbourne, 1988)\fR .sp 9p .RT .ce 0 .sp 1P .sp 2P .LP The\ CCITT, .sp 1P .RT .sp 1P .LP \fIconsidering\fR .sp 9p .RT .PP (a) that Recommendation X.1 specifies the international user classes of service in public data networks; .PP (b) that Recommendation X.2 specifies the international data transmission services and optional user facilities in public data networks; .PP (c) that Recommendation X.25 specifies the DTE/DCE interface for packet mode terminals connected to public data networks by dedicated circuit; .PP (d) that Recommendation X.75 specifies the packet\(hyswitched signalling system between public data networks providing data transmission services; .PP (e) that Recommendation X.323 specifies general arrangements for interworking between packet\(hyswitched public data networks; .PP ( f ) that Recommendation X.96 specifies call progress signals in public data networks; .PP (g) that Recommendation X.110 specifies the international routing principles and routing plan for public data networks; .PP (h) that Recommendation X.213 defines the OSI Network Layer service; .PP (i) that Recommendation X.140 defines general quality of service parameters for communication via public data networks; .PP ( j ) that Recommendation X.134 specifies portion boundaries and packet layer reference events for defining packet\(hyswitched performance parameters; .PP (k) that Recommendation X.135 specifies speed of service performance values for public data networks when providing international packet\(hyswitched service; .PP (l) that Recommendation X.137 specifies availability performance values for public data networks when providing international packet\(hyswitched service, .sp 1P .LP \fIunanimously declares\fR .sp 9p .RT .PP (1) that the accuracy and dependability parameters defined in this Recommendation shall be used in the planning and operation of international packet\(hyswitched data communication services provided in accordance with Recommendations\ X.25 and\ X.75; .PP (2) that in such services, the performance values specified in this Recommendation shall be taken as worst\(hycase limits under the conditions specified herein. .sp 2P .LP \fB1\fR \fBIntroduction\fR .sp 1P .RT .PP 1.1 This Recommendation is the third in a series of four CCITT Recommendations (X.134\(hyX.137) that define performance parameters and values for international packet\(hyswitched data communication services. Figure\ 1/X.136 illustrates the scope of these four Recommendations and the relationships among them. .bp .sp 9p .RT .LP .rs .sp 39P .ad r \fBFigure 1/X.136, (N), p.\fR .sp 1P .RT .ad b .RT .PP 1.2 Recommendation X.134 divides a virtual connection into basic sections whose boundaries are associated with X.25 and X.75 interfaces; defines particular collections of basic sections, called virtual connection portions, for which performance values will be specified; and defines a set of packet layer reference events (PEs) which provide a basis for performance parameter definition. The basic sections consist of network sections and circuit sections. They are delimited, in each case, by physical data terminal equipment (DTE) or data switching equipment (DSE) interfaces. Virtual connection portions are identified either as national portions or international portions. Each PE is defined to occur when a packet crossing a section boundary changes the state of the packet layer interface. .sp 9p .RT .PP 1.3 For comparability and completeness, packet\(hyswitched network performance is considered in the context of the 3\|\(mu\|3 performance matrix defined in Recommendation\ X.140. Three protocol\(hyindependent data communication functions are identified in the matrix: .bp .sp 9p .RT .LP access, user information transfer, and disengagement. These general functions correspond to call set\(hyup, data (and interrupt), transfer, and call clearing in packet\(hyswitched virtual call services conforming to the X.25 and X.75 Recommendations. Each function is considered with respect to three general performance concerns (or \*Qperformance criteria\*U): speed, accuracy, and dependability. These express, respectively, the delay or rate, degree of correctness, and degree of certainty with which the function is performed. .PP 1.4 Recommendation X.135 defines protocol\(hyspecific speed of service parameters and values associated with each of the three data communication functions. This Recommendation defines protocol\(hyspecific accuracy and dependability parameters and values associated with each function. The Recommendation\ X.135 and X.136 parameters are called \*Qprimary parameters\*U to emphasize their direct derivation from packet layer reference events. .sp 9p .RT .PP 1.5 An associated two\(hystate model provides a basis for describing overall service availability. A specified availability function compares the values for a subset of the primary parameters with corresponding outage thresholds to classify the service as \*Qavailable\*U (no service outage) or \*Qunavailable\*U (service outage) during scheduled service time. Recommendation\ X.137 specifies the availability function and defines the availability parameters and values that characterize the resulting binary random process. .sp 9p .RT .PP 1.6 Eight accuracy and dependability parameters are defined in this Recommendation: two access parameters (call set\(hyup error probability and call set\(hyup failure probability), five user information transfer parameters (residual error rate, reset stimulus probability, reset probability, premature disconnect stimulus probability, and premature disconnect probability), and one disengagement parameter (call clear failure probability). Each parameter can be applied to any basic section or portion of a virtual connection. This generally makes the parameters useful in performance allocation and concatenation. .sp 9p .RT .PP 1.7 This Recommendation specifies accuracy and dependability values for national and international portions of two types (Table\ 1/X.136). Performance values for data terminal equipment are not specified, but the parameters defined in this Recommendation may be employed in such specification to assist users in establishing quantitative relationships between network performance and quality of service (see Recommendation\ X.140). .sp 9p .RT .ce \fBH.T. [T1.136]\fR .ce TABLE\ 1/X.136 .ce \fBVirtual connection portion types for which\fR .ce \fBperformance values are specified\fR .ce \|\ua\d\u)\d .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(60p) | cw(120p) . Portion type Typical characteristics _ .T& lw(60p) | lw(120p) . National A T{ Terrestrial connection via an access network section T} _ .T& lw(60p) | lw(120p) . National B T{ Connection via an access network section with one satellite circuit; or via an access network section and one or more transit network sections T} _ .T& lw(60p) | lw(120p) . International A T{ Connection via a direct terrestrial internetwork section T} _ .T& lw(60p) | lw(120p) . International B T{ Connection via two satellite circuits and one transit network section; or via one satellite circuit and two or more transit network sections \ua\d\u)\d The values specified for Type B portions also apply to virtual connection portions not explicitly identified as Type A or Type B. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 1/X.136 [T1.136], p.\fR .sp 1P .RT .ad b .RT .LP .bp .PP 1.8 Worst\(hycase values for each of the eight accuracy and dependability parameters are specified below for each virtual connection portion type identified in Table\ 1/X.136. The term \*Qworst case\*U means that these values should be met during the normal busy hour in the worst\(hyperforming virtual connection portion used in providing international packet\(hyswitched services. The performance of a virtual connection portion may be better than the worst\(hycase values specified in this Recommendation. Design objectives that take into account more demanding user applications and network performance and connectivity enhancements are for further study. .sp 9p .RT .PP Numerical methods for combining individual portion performance values to estimate end\(hyto\(hyend performance are also provided in this Recommendation. DTE to DTE values for two particular hypothetical reference connections are derived using these methods in Annex\ B. .sp 2P .LP \fB2\fR \fBAccess parameters\fR .sp 1P .RT .PP This section specifies worst\(hycase values for two access parameters: call set\(hyup error probability and call set\(hyup failure probability. .PP Call set\(hyup error and call set\(hyup failure are defined between pairs of section boundaries (\fIB\fR\d\fIi\fR\u, \fIB\fR\d\fIj\fR\u). \fIB\fR\d\fIj\fR\uis one of the set of boundaries to which the call attempt can properly be routed. Figure\ 2/X.136 identifies the sequence of four particular events that occur at these boundaries during a successful call set up .FS The PE numbers in Figure\ 2a/X.136 refer to Tables 1 and 2 in Recommendation\ X.134. .FE . A call set\(hyup attempt over this section is an occurrence of event\ (a). A successful call set\(hyup attempt over this section is a sequential occurrence of corresponding events (a, b, c and\ d) within a 200\(hysecond timeout period .FS This period corresponds to timer T21 in Recommendation X.25. .FE . Call set\(hyup errors and call set\(hyup failures within this section are defined below. Any other unsuccessful call set\(hyup attempt is caused by problems outside the section and is excluded from the measurement. .RT .sp 1P .LP 2.1 \fICall set\(hyup error probability\fR .sp 9p .RT .PP Call set\(hyup error probability applies to virtual call services. It does not apply to permanent virtual circuit establishment. This parameter is used to measure the accuracy of the general user function of access in public packet\(hyswitched services conforming to Recommendations\ X.25 and\ X.75. .RT .sp 1P .LP 2.1.1 \fBcall set\(hyup error probability\fR \fIdefinition\fR .sp 9p .RT .PP Call set\(hyup error probability is the ratio of total call attempts that result in call set\(hyup error to the total call attempts in a population of interest. .PP With reference to Figure 2/X.136, a call set\(hyup error is defined to occur on any call attempt in which event\ (d) occurs, but event\ (c) does not occur within a 200\(hysecond timeout period. .RT .PP Call set\(hyup error is essentially the case of a \*Qwrong number\*U. It occurs when the network responds to a valid call request by erroneously establishing a virtual call to a destination DTE other than the one designated in the call request, and does not correct the error prior to entry to the data transfer state. It may be caused, for example, by network operator administrative or maintenance actions. .RT .LP .sp 1 .bp .ce \fBH.T. [T2.136]\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(48p) | cw(24p) sw(24p) | cw(24p) sw(24p) , ^ | c | c | c | c. Boundary/Event Interface \fIB\fI \fIB\fI (a) (d) (b) (c) _ .T& cw(48p) | cw(24p) | cw(24p) | cw(24p) | cw(24p) . X.25 2 3 1 4 _ .T& cw(48p) | cw(24p) | cw(24p) | cw(24p) | cw(24p) . X.75 1 2 1 T{ 2 \fIa)\ Packet layer reference events (PEs)\fR T} _ .TE .nr PS 9 .RT .ad r \fBTableau (avec la figure 2/X.136) [T2.136], p.21\fR .sp 1P .RT .ad b .RT .LP .rs .sp 23P .ad r \fBFigure 2/X.136 (avec le tableau), (N), p.22\fR .sp 1P .RT .ad b .RT .PP Call set\(hyup error is distinguished from successful call set\(hyup by the fact that the intended called user is not contacted and committed to the data communication session during the call set\(hyup attempt. .PP Call set\(hyup error probability does not apply to the fast select mode of data transfer. The optional user call redirection facilities in X.25 (including hunt group, call redirection, call forwarding subscription, call forwarding selection, call redirection or forwarding notification, and called line address modified notification) are assumed not to be used in the calculation of this parameter. .PP The specific X.134 PEs used in measuring call set\(hyup error probability at each section boundary are those identified in Figure\ 2/X.136. .bp .RT .sp 1P .LP 2.1.2 \fIValues\fR .sp 9p .RT .PP The contribution from each network portion to the overall call set\(hyup error probability under the conditions described in this Recommendation shall not exceed the values specified in Table\ 2/X.136. .RT .ce \fBH.T. [T3.136]\fR .ce TABLE\ 2/X.136 .ce \fBWorst\(hycase call set\(hyup error probability values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . Probability 10\uD\dlF261\u5\d 2 \(mu 10\uD\dlF261\u5\d \ua\d\u)\d T{ 2 \(mu 10\uD\dlF261\u5\d \ua\d\u)\d The Type A international virtual connection portion consists only of a physical circuit. Its contribution to call set\(hyup error probability is expected to be negligible. .parag \fINote\fR \ \(em\ All specified values are provisional. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 2/X.136 [T3.136], p.\fR .sp 1P .RT .ad b .RT .LP .sp 2 .sp 1P .LP 2.2 \fICall set\(hyup failure probability\fR .sp 9p .RT .PP Call set\(hyup failure probability applies only to virtual call services. This parameter is used to measure the dependability of the general user function of access in public packet\(hyswitched services conforming to Recommendations\ X.25 and\ X.75. .RT .sp 1P .LP 2.2.1 \fBcall set\(hyup failure probability\fR \fIdefinition\fR .sp 9p .RT .PP Call set\(hyup failure probability is the ratio of total call attempts that result in call set\(hyup failure to the total call attempts in a population of interest. .PP With reference to Figure 2/X.136, call set\(hyup failure is defined to occur on any call attempt in which either one of the following outcomes is observed within a 200\(hysecond timeout period .FS Recommendation X.96 places limits on the frequency at which a DTE can repeat call attempts to a given destination. .FE : .RT .LP 1) Both events (b) and (d) do not occur. .LP 2) Events (b) and (c) occur, but event (d) does not. .PP Call attempts that are cleared by the section as a result of incorrect performance or nonperformance on the part of an entity outside the section are excluded. The specific X.134 PEs used in measuring call set\(hyup failure probability at each section boundary are those identified in Figure\ 2/X.136. .bp .sp 1P .LP 2.2.2 \fIValues\fR .sp 9p .RT .PP The contribution from each network portion to the overall call set\(hyup failure probability under the conditions described in this Recommendation shall not exceed the values specified in Table\ 3/X.136. .RT .ce \fBH.T. [T4.136]\fR .ce TABLE\ 3/X.136 .ce \fBWorst\(hycase call set\(hyup failure probability values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . Probability 5 \(mu 10\uD\dlF261\u3\d 10\uD\dlF261\u2\d \ua\d\u)\d T{ 10\uD\dlF261\u2\d \ua\d\u)\d The Type A international virtual connection portion consists only of a physical circuit. Its contribution to call set\(hyup failure probability is expected to be negligible. .parag \fINote\fR \ \(em\ All specified values are provisional. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 3/X.136 [T4.136], p.\fR .sp 1P .RT .ad b .RT .sp 1P .LP 2.2.3 \fIExcluded call attempts\fR .sp 9p .RT .PP A call set\(hyup attempt can also fail as a result of user blocking. Such failures are excluded from network performance measurement. Examples of user blocking include the following: .RT .LP 1) Either the originating or the called user issues a clear request to reject the call set\(hyup attempt. .LP 2) The called user delays excessively in generating the call accepted packet during the connection period, with the result that a connection is not established before the timeout. .LP 3) All logical channels at the called DTE are in use. .sp 2P .LP \fB3\fR \fBUser information transfer parameters\fR .sp 1P .RT .PP This section specifies worst\(hycase values for five user information transfer parameters; residual error rate, reset stimulus probability, reset probability, premature disconnect stimulus probability, and premature disconnect probability. These parameters describe impairments observed during the data transfer state of a virtual call or permanent virtual circuit. .RT .sp 1P .LP 3.1 \fIResidual error rate\fR .sp 9p .RT .PP Residual error rate applies to both virtual call and permanent virtual circuit services. This parameter is used to measure the accuracy of the general function of user information transfer in public packet\(hyswitched services conforming to Recommendations\ X.25 and\ X.75. .RT .sp 1P .LP 3.1.1 \fBresidual error rate\fR \fIdefinition\fR .sp 9p .RT .PP Residual error rate is the ratio of total incorrect, lost, and extra (e.g.\ duplicate) user data bits to total user data bits transferred across either section boundary in a population of interest. .bp .PP User data bits are the bits of the user data field in data packets of the X.25 or X.75 packet layer (protocols and data above the packet layer). Framing routing, bit stuffing, error control, and other protocol fields introduced by all protocols at or below the packet layer are excluded. .PP Relationships among the quantities identified above are defined in Figure\ 3/X.136. Incorrect user data bits are user data bits that are inverted in transfer between the section boundaries, i.e.,\ bits whose binary value observed at the section boundary on the destination side of a virtual connection portion is the opposite of that observed at the section boundary on the source side. Lost user data bits are user data bits that are transferred into a virtual connection portion at one section boundary, but are not transferred out of the virtual connection portion at the other within 200\ seconds of non\(hyflow\(hycontrolled transmission. Bits lost in association with a reset or premature disconnect are excluded in calculating residual error rate. Extra user data bits are user data bits that are transferred out of a virtual connection portion at one section boundary, but were not previously transferred into the virtual connection portion at the other. Extra user data bits include duplicated user data bits and misdelivered user data bits. .RT .LP .rs .sp 20P .ad r \fBFigure 3/X.136, (N), p.\fR .sp 1P .RT .ad b .RT .PP The specific X.134 PEs used in measuring residual error rate at each section boundary are identified in Table\ 4/X.136. Only user data bits in data packets that create the specified PEs are counted in calculating residual error rate estimates. .ce \fBH.T. [T5.136]\fR .ce TABLE\ 4/X.136 .ce \fBPacket layer reference events (PEs) used\fR .ce \fBin measuring residual error rate\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(84p) | cw(48p) . Circuit section Starting/Ending PE _ .T& lw(84p) | cw(48p) . Source access circuit section 10a (X.25) _ .T& lw(84p) | cw(48p) . T{ Destination access circuit section T} 9a (X.25) _ .T& lw(84p) | cw(48p) . Internetwork circuit section 5a (X.75) _ .TE .nr PS 9 .RT .ad r \fBTableau 4/X.136 [T5.136], p.\fR .sp 1P .RT .ad b .RT .LP .bp .PP In practice, it is not possible in all cases to distinguish lost, errored, and extra bit occurrences without detailed knowledge of the problems within the boundaries. A simple, approximate method of calculating residual error rate values is presented in Annex\ A. Other methods of equivalent or superior accuracy are acceptable. .sp 1P .LP 3.1.2 \fIValues\fR .sp 9p .RT .PP The contribution from each network portion to the overall residual error rate of a virtual connection provided under the conditions described in this Recommendation shall not exceed the values specified in Table\ 5/X.136. This specified values are based on an assumed data packet length of 128\ octets. .RT .ce \fBH.T. [T6.136]\fR .ce TABLE\ 5/X.136 .ce \fBWorst\(hycase residual error rate values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . Probability 10\uD\dlF261\u1\d\u0\d 2 \(mu 10\uD\dlF261\u1\d\u0\d \ua\d\u)\d T{ 2 \(mu 10\uD\dlF261\u1\d\u0\d \ua\d\u)\d The Type A international virtual connection portion consists only of a physical circuit. Its contribution to residual error rate is expected to be negligible. .parag \fINote\fR \ \(em\ All specified values are provisional. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 5/X.136 [T6.136], p.\fR .sp 1P .RT .ad b .RT .sp 1P .LP 3.1.3 \fIComponents of residual error rate\fR .sp 9p .RT .PP In some applications, it may be important to specify probability limits for the individual failure outcomes illustrated in Figure\ 3/X.136 in addition to the overall residual error rate. The general user information error, user information loss, and extra user information delivery probabilities defined in Recommendation\ X.140 may be specialized to the corresponding user data bit\(hyoriented measures as follows. .RT .LP \(em User data bit error probability \fIP\fR\d1\u(\fIE\fR ) is the ratio of total incorrect user data bits (\fIN\fR\d\fIE\fR\u) to total successfully transferred user data bits \fIplus\fR incorrect user data bits (\fIN\fR\d\fIS\fR\u\ +\ \fIN\fR\d\fIE\fR\u) in a population of interest. .LP \(em User data bit loss probability \fIP\fR\d1\u(\fIL\fR ) is the ratio of total lost user data bits (\fIN\fR\d\fIL\fR\u) to total transmitted user data bits (\fIN\fR\d\fIT\fR\u) in a population of interest. .LP \(em Extra user data bit delivery probability \fIP\fR\d1\u(\fIX\fR ) is the ratio of total (unrequested) extra user data bits (\fIN\fR\d\fIX\fR\u) to total received user data bits (\fIN\fR\d\fIR\fR\u) in a population of interest. .PP The denominators of these ratios are chosen to ensure that each defined probability is properly normalized; i.e.,\ each failure outcome is expressed in proportion to the total number of opportunities for that outcome to occur. The mathematical relationship between residual error rate (\fIRER\fR ) and the three user data bits transfer failure probabilities defined above is as follows. \v'6p' .sp 1P .ce 1000 \fIRER\fR = {[\fIP\fR\d1\u(\fIE\fR )] [\fIN\fR\d\fIE\fR\u+ \fIN\fR\d\fIS\fR\u] + [\fIP\fR\d1\u(\fIL\fR )] [\fIN\fR\d\fIT\fR\u] + [\fIP\fR\d1\u(\fIX\fR )] [\fIN\fR\d\fIR\fR\u] } over {\fIN\fR } .ce 0 .sp 1P .LP .sp 1 .bp .sp 1P .LP 3.2 \fIReset parameters\fR .sp 9p .RT .PP Reset stimulus probability and reset probability are related parameters used to describe the dependability of the general function of user information transfer in public packet\(hyswitched services conforming to Recommendations\ X.25 and\ X.75. .RT .sp 1P .LP 3.2.1 \fBreset stimulus probability\fR \fIdefinition\fR .sp 9p .RT .PP A reset stimulus is observed at a single section boundary. It is any event or combination of events that according to the protocol should result in a reset (or, in the case of a PVC, a reset or restart) being generated by the recipient .FS For the purpose of performance parameter definition it is assumed that the reset stimuli for an X.25\ DTE are equivalent to the reset stimuli for an X.25 DCE. .FE . An example of a reset stimulus is a DTE transmitting a reject packet when the packet retransmission facility has not been subscribed. .PP The \fBreset stimulus probability of a section at a boundary\fR is the expected number of reset stimuli generated within that section and transferred across the boundary per virtual connection second. .RT .sp 1P .LP 3.2.2 \fBreset probability\fR \fIdefinition\fR .sp 9p .RT .PP A reset event is defined to have been generated within a section when, in the absence of an external reset stimulus, two packets exit the section \(em\ one at each boundary\ \(em creating any one of the pairs of Recommendation\ X.134 packet layer reference events listed in Table\ 6/X.136. .RT .LP .sp 1 .ce \fBH.T. [T7.136]\fR .ce TABLE\ 6/X.136 .ce \fBPacket layer reference events (PEs) used in\fR .ce \fBmeasuring reset probability\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(60p) | cw(90p) . Boundaries of section Pair of PEs _ .T& cw(60p) | lw(90p) . X.25\ X.25 [20(X.25)\ 20(X.25)] .T& cw(60p) | lw(90p) . X.25\ X.75 [20(X.25)\ 10(X.75)] .T& cw(60p) | lw(90p) . X.75\ X.75 [10(X.75)\ 10(X.75)] _ .T& cw(150p) . T{ \fIa)\ Pairs of PEs resulting from reset events\fR T} .T& cw(60p) | cw(90p) . Boundaries of section Pair of PEs _ .T& cw(60p) | lw(90p) . X.25\ X.25 [20(X.25)\ 24(X.25)] .T& cw(60p) | lw(90p) . X.25\ X.75 T{ [20(X.25)\ 12(X.75)] or [24(X.25)\ 10(X.75)] T} .T& cw(60p) | lw(90p) . X.75\ X.75 [10(X.75)\ 12(X.75)] _ .T& cw(150p) . T{ \fIb)\ Additional PE pairs resulting from reset events on PVCs\fR T} .TE .nr PS 9 .RT .ad r \fBTableau 6/X.136 [T7.136], p.\fR .sp 1P .RT .ad b .RT .LP .sp 4 .bp .PP The reset probability for a virtual connection section is the probability, in any given second, that a reset event is generated within that section. .PP Reset events generated within a section may be estimated by counting the number of reset request and reset indication packets exiting the section during a measurement period; subtracting the number of reset request and reset indication packets entering the section during the same period; dividing the difference by\ 2; and then substracting from the result any reset stimuli that enter the section during the period. .PP \fINote\fR \ \(em\ Reset events may be associated with a loss of packets. .PP The specific X.134 PEs used in measuring reset probability at each section boundary are identified in Table\ 6/X.136. .RT .sp 1P .LP 3.2.3 \fIValues\fR .sp 9p .RT .PP The contribution from each network portion to overall reset stimulus probability and reset probability under the conditions described in this Recommendation shall not exceed the values specified in Table\ 7/X.136. .RT .ce \fBH.T. [T8.136]\fR .ce TABLE\ 7/X.136 .ce \fBWorst\(hycase reset stimulus probability and reset\fR .ce \fBprobability values for virtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(72p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(72p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . T{ Reset stimulus probability (reset stimuli per VC second) T} 10\uD\dlF261\u6\d 10\uD\dlF261\u6\d \ua\d\u)\d 10\uD\dlF261\u6\d _ .T& lw(72p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . T{ Reset probability (resets per VC second) T} 10\uD\dlF261\u5\d 2 \(mu 10\uD\dlF261\u5\d N/A T{ 2 \(mu 10\uD\dlF261\u5\d \ua\d\u)\d The Type A international virtual connection portion consists only of a physical circuit. Its contribution to reset stimulus probability is expected to be negligible. .parag \fINote\fR \ \(em\ All specified values are provisional. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 7/X.136 [T8.136], p.\fR .sp 1P .RT .ad b .RT .PP The reset stimulus and reset probabilities for a set of concatenated virtual connection portions may be estimated from the individual portion probabilities as follows. Assume between boundaries (\fIB\fR\d\fIi\fR\u, \fIB\fR\d\fIj\fR\u) the reset probability is \fIR\fR\d1\uand the reset stimulus probabilities are \fIS\fR\d1\u\fI\fI\d\fIi\fR\u, \fIS\fR\d1\u\fI\fI\d\fIj\fR\u. Assume between boundaries (\fIB\fR\d\fIj\fR\u, \fIB\fR\d\fIk\fR\u) the reset probability is \fIR\fR\d2\uand the reset stimulus probabilities are \fIS\fR\d2\u\fI\fI\d\fIj\fR\u, \fIS\fR\d2\u\fI\fI\d\fIk\fR\u. Then on a VC passing through \fIB\fR\d\fIj\fR\uthe reset probability between \fIB\fR\d\fIi\fR\uand \fIB\fR\d\fIk\fR\uis approximately (\fIR\fR\d1\u\ +\fIR\fR\d2\u\ + \fIS\fR\d1\u\fI\fI\d\fIj\fR\u\ + \fIS\fR\d2\u\fI\fI\d\fIj\fR\u). See Figure\ 4/X.136. The reset stimulus probability at \fIB\fR\d\fIi\fR\uis \fIS\fR\d1\u\fI\fI\d\fIi\fR\uand the reset stimulus probability at \fIB\fR\d\fIk\fR\uis \fIS\fR\d2\u\fI\fI\d\fIk\fR\u. .sp 1P .LP 3.3 \fIPremature disconnect parameters\fR .sp 9p .RT .PP Premature disconnect stimulus probability and premature disconnect probability are related parameters used to describe the dependability of user information transfer in public packet\(hyswitched networks conforming to Recommendations\ X.25 and\ X.75. .bp .RT .LP .rs .sp 14P .ad r \fBFigure 4/X.136, (N), p.30\fR .sp 1P .RT .ad b .RT .sp 1P .LP 3.3.1 \fBpremature disconnect stimulus probability\fR \fIdefinition\fR .sp 9p .RT .PP A premature disconnect stimulus is observed at a single section boundary. It is any event or combination of events that according to the protocol should result in a clear or restart being generated by the recipient .FS For the purpose of performance parameter definition, it is assumed that the premature disconnect stimuli for an X.25 DTE are equivalent to the premature disconnect stimuli for an X.25\ DCE. .FE . An example of of a premature disconnect stimulus is the transmission of an incorrect packet type into a virtual connection section. A premature disconnect stimulus applies only to virtual call services. .PP The \fBpremature disconnect stimulus probability of a section at a\fR \fBboundary\fR is the probability of a premature disconnect stimulus generated within that section and transferred across the boundary per virtual connection second. .RT .sp 1P .LP 3.3.2 \fBpremature disconnect probability\fR \fIdefinition\fR .sp 9p .RT .PP A premature disconnect event is defined to have been generated within a section when, in the absence of an external premature disconnect stimulus, two packet exit the section \(em\ one at each boundary\ \(em creating any one of the pairs of packet layer reference events listed in Table\ 8/X.136. A premature disconnect event applies only to virtual call services. .RT .ce \fBH.T. [T9.136]\fR .ce TABLE\ 8/X.136 .ce \fBPacket layer reference events (PEs) used in\fR .ce \fBmeasuring premature disconnect probability\fR .ce (Pairs of PEs resulting from premature disconnect events) .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(60p) | cw(120p) . Boundaries of section Pair of PEs _ .T& cw(60p) | lw(120p) . X.25\ X.25 T{ [5(X.25)\ 5(X.25)] or [5(X.25)\ 24(X.25)] T} .T& cw(60p) | lw(120p) . X.25\ X.75 T{ [5(X.25)\ 3(X.75)] or [5(X.25)\ 12(X.75)] or [24(X.25)\ 3(X.75)] T} .T& cw(60p) | lw(120p) . X.75\ X.75 T{ [3(X.75)\ 3(X.75)] or [3(X.75)\ 12(X.75)] T} _ .TE .nr PS 9 .RT .ad r \fBTableau 8/X.136 [T9.136], p.\fR .sp 1P .RT .ad b .RT .LP .bp .PP The premature disconnect probability for a virtual connection section is the probability, in any given second, that a virtual call experiences a premature disconnect event generated within that section. .PP Premature disconnect events generated within a section may be estimated by counting the number of clear request or clear indication packets exiting the section during a measurement period; subtracting the number of clear request and clear indication packets entering the section during the same period; dividing the difference by two; and then substracting from the result any premature disconnect stimuli that enter the section during that period. .PP \fINote\fR \ \(em\ Premature disconnect events may be associated with a loss of packets. .PP The specific X.134 PEs used in measuring premature disconnect probability at each section boundary are identified in Table\ 8/X.136. .RT .sp 1P .LP 3.3.3 \fIValues\fR .sp 9p .RT .PP The contribution from each network portion to overall premature disconnect stimulus probability and premature disconnect probability under the conditions described in this Recommendation shall not exceed the values specified in Table\ 9/X.136. .RT .ce \fBH.T. [T10.136]\fR .ce TABLE\ 9/X.136 .ce \fBWorst\(hycase premature disconnect stimulus probability\fR .ce \fBand premature disconnect probability values for\fR .ce \fBvirtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(90p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(90p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . T{ Premature disconnect stimulus probability (premature disconnect stimuli per VC second) T} 10\uD\dlF261\u7\d 10\uD\dlF261\u7\d 10\uD\dlF261\u7\d 10\uD\dlF261\u7\d _ .T& lw(90p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . T{ Premature disconnect probability (premature disconnects per VC second) T} 5 \(mu 10\uD\dlF261\u6\d 10\uD\dlF261\u5\d N/A T{ 10\uD\dlF261\u5\d \fINote\fR \ \(em\ All specified values are provisional. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 9/X.136 [T10.136], p.\fR .sp 1P .RT .ad b .RT .PP The premature disconnect stimulus and premature disconnect probabilities for a set of concatenated virtual connection portions may be estimated from the individual portion probabilities in a manner analogous to that described in \(sc\ 3.2.3. .sp 2P .LP \fB4\fR \fBDisengagement performance \(em call clear failure probability\fR .sp 1P .RT .PP Call clear failure probability applies only to virtual call services. This parameter is used to measure the accuracy and dependability of the general function of disengagement in public packet\(hyswitched services conforming to Recommendations\ X.25 and\ X.75. .RT .sp 1P .LP 4.1 \fBcall clear failure probability\fR \fIdefinition\fR .sp 9p .RT .PP Call clear failure is defined with reference to events at the boundaries of a virtual connection section (\fIB\fR\d\fIi\fR\u, \fIB\fR\fI\d\fIj\fR\u). A call clear attempt occurs when a clean request or clear indication packet enters the section creating a packet layer reference event at \fIB\fR\d\fIi\fR\u. A call clear failure occurs when no corresponding clear indication packet layer reference event occurs at \fIB\fR\d\fIj\fR\uwithin 180\ seconds. The relevant PEs are listed in Table\ 10/X.136. .bp .RT .ce \fBH.T. [T11.136]\fR .ce TABLE\ 10/X.136 .ce \fBPacket layer reference events (PEs) used in\fR .ce \fBmeasuring call clear failure probability\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(90p) | cw(36p) sw(36p) , ^ | c | c. Circuit section T{ X.134 Packet layer reference event T} Starting PE Ending PE _ .T& lw(90p) | cw(36p) | cw(36p) . T{ Clearing DTE access circuit section T} 6(X.25) \(em _ .T& lw(90p) | cw(36p) | cw(36p) . T{ Cleared DTE access circuit section T} \(em 5(X.25) (does not occur) _ .T& lw(90p) | cw(36p) | cw(36p) . Internetwork circuit section 3(X.75) 3(X.75) (does not occur) _ .TE .nr PS 9 .RT .ad r \fBTableau 10/X.136 [T11.136], p.\fR .sp 1P .RT .ad b .RT .PP Call clear failure probability for a virtual connection section is the ratio of call clear failures to call clear atempts in a population of interest. .sp 1P .LP 4.2 \fIValues\fR .sp 9p .RT .PP The contribution from each virtual connection portion to the overall call clear failure probability under the conditions described in this Recommendation shall not exceed the values specified in Table\ 11/X.136. .RT .ce \fBH.T. [T12.136]\fR .ce TABLE\ 11/X.136 .ce \fBWorst\(hycase call clear failure probability values\fR .ce \fBfor virtual connection portions\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(42p) | cw(30p) sw(30p) sw(30p) sw(30p) , ^ | c s | ^ , ^ | c | c | c s ^ | ^ | ^ | c | c. Statistic T{ Virtual connection portion type T} National A B International A B _ .T& lw(42p) | cw(30p) | cw(30p) | cw(30p) | cw(30p) . Probability 10\uD\dlF261\u5\d 2 \(mu 10\uD\dlF261\u5\d \ua\d\u)\d T{ 2 \(mu 10\uD\dlF261\u5\d \ua\d\u)\d The Type A international virtual connection portion consists only of a physical circuit. Its contribution to call clear failure probability is expected to be negligible. .parag \fINote\fR \ \(em\ All specified values are provisional. .parag T} _ .TE .nr PS 9 .RT .ad r \fBTableau 11/X.136 [T12.136], p.\fR .sp 1P .RT .ad b .RT .sp 1P .LP 4.3 \fILocal clear confirmation\fR .sp 9p .RT .PP The failure of a section to respond to a clear request or clear indication packet with a clear confirmation packet is not addressed in this Recommendation. Recovery mechanisms for such occurrences are defined in both the Recommendations\ X.25 and X.75 protocols. Clear confirmation at X.25 interfaces is a national matter. .bp .RT .ce 1000 ANNEX\ A .ce 0 .ce 1000 (to Recommendation X.136) .sp 9p .RT .ce 0 .ce 1000 \fBAncilliary information on \fR \fBaccurancy and\fR .sp 1P .RT .ce 0 .sp 1P .ce 1000 \fBdependability measurement\fR .ce 0 .sp 1P .PP The following points should be noted with regard to accuracy and dependability measurement: .sp 1P .RT .LP \(em The ratios used to calculate the probabilities are understood to be estimates of the true probabilities. .LP \(em The periods of observation for accuracy \(em and dependability \(em related probabilities, as well as the concept of busy hour itself, for packet services, are for further study. .PP Figure A1/X.136 illustrates a simple approximate method of calculating residual error rate. A sample consisting of \fIn\fR\d\fIT\fR\uuser data bits is transmitted typically in many successive packets. (A 128\(hyoctet packet length is assumed.) A corresponding sample consisting of \fIn\fR\d\fIR\fR\uuser data bits is received. If \fIn\fR\d\fIT\fR\u\ =\ \fIn\fR\d\fIR\fR\u, the transmitted and received user data bits are compared bit for bit, and the number of incorrect data bits in the sample is estimated by \fIm\fR\d\fIE\fR\u, the number of corresponding transmitted and received bits that do not match. If \fIn\fR\d\fIT\fR\u\ >\ \fIn\fR\d\fIR\fR\u, the number of lost data bits in the sample is estimated by \fIm\fR\d\fIL\fR\u\ =\ (\fIn\fR\d\fIT\fR\u\ \(em\ \fIn\fR\d\fIR\fR\u). If \fIn\fR\d\fIT\fR\u\ <\ \fIn\fR\d\fIR\fR\u, the number of extra data bits in the sample is estimated by \fIm\fR\d\fIx\fR\u\ =\ (\fIn\fR\d\fIR\fR\u\ \(em\ \fIn\fR\d\fIT\fR\u). If a reset request or clear request is issued during the transfer of a measurement sample, that sample is excluded in calculating the RER estimate. .PP The outcome totals in each sample are accumulated over a number of samples sufficient to calculate the residual error rate with the desired precision. Guidelines for relating overall sample size with desired precision are for further study. It should be noted that the approximate method of residual error rate estimation presented here will not produce unbiased estimates if more than one category of bit transfer failure occurs in the same sample. Other, more exact methods of estimating residual error rate may also be employed. .RT .LP .rs .sp 01P .ad r BLANC .ad b .RT .LP .bp .LP .rs .sp 40P .ad r \fBFigure A\(hy1/X.136, (N), p.35\fR .sp 1P .RT .ad b .RT .LP .bp .ce 1000 ANNEX\ B .ce 0 .ce 1000 (to Recommendation X.136) .sp 9p .RT .ce 0 .ce 1000 \fBRepresentative end\(hyto\(hyend accuracy and\fR .sp 1P .RT .ce 0 .sp 1P .ce 1000 \fBdependability performance\fR .ce 0 .sp 1P .PP This annex provides two examples to illustrate how end\(hyto\(hyend (DTE to DTE) accuracy and dependability performance can be estimated from the individual virtual connection portion performance values specified in Recommendation\ X.136. Two example concatenations of Type\ A and Type\ B virtual connection portions are defined. The end\(hyto\(hyend call set\(hyup error probability, call set\(hyup failure probability, residual error rate, reset stimulus probability, reset probability, premature disconnect stimulus probability, premature disconnect probability, and call clear failure probability are calculated for each example. Although alternative network models and statistical assumptions are possible, the methods presented in this annex provide one practical way of estimating end\(hyto\(hyend performance from the performance of the individual network portions. .sp 1P .RT .sp 1P .LP B.1 \fIDefinition of the example end\(hyto\(hyend connections\fR .sp 9p .RT .PP For ease of reference, the two example end\(hyto\(hyend (i.e., DTE to DTE) connections presented in this annex will be referred to as \*QType\ 1\*U and \*QType\ 2\*U configurations. These hypothetical, but representative, configurations use the portion boundaries and packet layer reference events described in Recommendation\ X.134. Table\ 1/X.136 defines the virtual connection portion Types. .PP The Type 1 configuration is defined to be: .RT .ce \fBH.T. [T13.136]\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(24p) | lw(108p) | lw(24p) . .T& cw(24p) | lw(108p) | cw(24p) . DTE DTE .T& lw(42p) | lw(48p) | lw(42p) . .TE .nr PS 9 .RT .ad r \fBTableau [T13.136], p.\fR .sp 1P .RT .ad b .RT .PP The Type 2 configuration is defined to be: .ce \fBH.T. [T14.136]\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; lw(24p) | lw(108p) | lw(24p) . .T& cw(24p) | lw(108p) | cw(24p) . DTE DTE .T& lw(42p) | lw(48p) | lw(42p) . .TE .nr PS 9 .RT .ad r \fBTableau [T14.136], p.\fR .sp 1P .RT .ad b .RT .sp 1P .LP B.2 \fIEnd\(hyto\(hyend accuracy and dependability performance for the Type\ 1\fR \fIand Type\ 2 configuration examples\fR .sp 9p .RT .PP End\(hyto\(hyend accuracy and dependability performance values have been calculated for the example Type\ 1 and Type\ 2 connection configurations and are reported below in Tables\ B\(hy1/X.136 and\ B\(hy2/X.136. These calculations have been made by applying the methods described below to the individual network portions that, for convenience in defining these examples, are characterized by the worst\(hycase accuracy and dependability performance values specified in Recommendation\ X.136. .bp .PP Assuming that the performance associated with the individual network portions are statistically independent, then a very close approximation to the end\(hyto\(hyend performance can be obtained for the call set\(hyup error probability, call set\(hyup failure probability, residual error rate probability, and call clear failure probability by simply summing the respective probabilities for the concatenated individual connection portions. Note that this procedure assumes that the approximation error caused by dropping the higher order terms in combining these individual portion probabilities is usually not significant and therefore can be ignored for most cases of practical interest. .RT .LP \fIExample:\fR \ To compute the end\(hyto\(hyend probability of call set\(hyup error for the Type\ 1 configuration, refer to Table\ 2/X.136 to obtain the individual portion probabilities (National\ A: probability\ =\ 10\uD\dlF261\u5\d; International\ A: probability\ =\ 0). The end\(hyto\(hyend probability of call set\(hyup error is then 10\uD\dlF261\u5\d\ + 0\ +\ 10\uD\dlF261\u5\d\ = 2\ * \ 10\uD\dlF261\u5\d. .PP The approximate end\(hyto\(hyend performance at each boundary for the reset probability, reset stimulus probability, premature disconnect stimulus probability, and premature disconnect probability can be calculated using methods in \(sc\(sc\ 3.2.3 and\ 3.3.3 of Recommendation\ X.136. .LP \fIExample:\fR \ To compute the end\(hyto\(hyend performance for the reset probability for the Type\ 2 configuration, refer to Table\ 7/X.136 to obtain the individual portion probabilities. The end\(hyto\(hyend probability of reset at the boundaries can be calculated as 10\uD\dlF261\u5\d\ + 0\ +10\uD\dlF261\u5\d\ + 0\ +10\uD\dlF261\u6\d\ + 0\ +10\uD\dlF261\u6\d\ = 2.2\ * \ 10\uD\dlF261\u5\d. .LP \fIExample:\fR \ To compute the end\(hyto\(hyend performance for the reset stimulus probability for the Type\ 1 configuration, refer to Table\ 7/X.136 to obtain the individual portion probabilities. The end\(hyto\(hyend probability of reset stimulus at the boundaries can be determined by inspection as 10\uD\dlF261\u6\d. .LP .sp 1 .ce \fBH.T. [T15.136]\fR .ce TABLE\ B\(hy1/X.136 .ce \fBEnd\(hyto\(hyend accuracy and dependability performance\fR .ce \fBfor the Type 1 configuration example\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(156p) . Type 1 configuration _ .T& cw(120p) | cw(36p) . Statistic End\(hyto\(hyend value _ .T& lw(120p) | lw(36p) . T{ Call set\(hyup error probability T} 2 * 10\uD\dlF261\u5\d _ .T& lw(120p) | lw(36p) . T{ Call set\(hyup failure probability T} 1 * 10\uD\dlF261\u2\d _ .T& lw(120p) | lw(36p) . Residual error rate 2 * 10\uD\dlF261\u1\d\u0\d _ .T& lw(120p) | lw(36p) . Reset stimulus probability 1 * 10\uD\dlF261\u6\d _ .T& lw(120p) | lw(36p) . Reset probability 2.2 * 10\uD\dlF261\u5\d _ .T& lw(120p) | lw(36p) . T{ Premature disconnect stimulus probability T} 1 * 10\uD\dlF261\u7\d _ .T& lw(120p) | lw(36p) . T{ Premature disconnect probability T} 1.04 * 10\uD\dlF261\u5\d _ .T& lw(120p) | lw(36p) . T{ Call clear failure probability T} 2 * 10\uD\dlF261\u5\d _ .TE .nr PS 9 .RT .ad r \fBTableau B\(hy1/X.136 [T15.136], p.\fR .sp 1P .RT .ad b .RT .LP .bp .ce \fBH.T. [T16.136]\fR .ce TABLE\ B\(hy2/X.136 .ce \fBEnd\(hyto\(hyend accuracy and dependability performance\fR .ce \fBfor the Type 2 configuration example\fR .ps 9 .vs 11 .nr VS 11 .nr PS 9 .TS center box; cw(156p) . Type 2 configuration _ .T& cw(120p) | cw(36p) . Statistic End\(hyto\(hyend value _ .T& lw(120p) | lw(36p) . T{ Call set\(hyup error probability T} 5 * 10\uD\dlF261\u5\d _ .T& lw(120p) | lw(36p) . T{ Call set\(hyup failure probability T} 2.5 * 10\uD\dlF261\u2\d _ .T& lw(120p) | lw(36p) . Residual error rate 5 * 10\uD\dlF261\u1\d\u0\d _ .T& lw(120p) | lw(36p) . Reset stimulus probability 1 * 10\uD\dlF261\u6\d _ .T& lw(120p) | lw(36p) . Reset probability 5.4 * 10\uD\dlF261\u5\d _ .T& lw(120p) | lw(36p) . T{ Premature disconnect stimulus probability T} 1 * 10\uD\dlF261\u7\d _ .T& lw(120p) | lw(36p) . T{ Premature disconnect probability T} 2.54 * 10\uD\dlF261\u5\d _ .T& lw(120p) | lw(36p) . T{ Call clear failure probability T} 5 * 10\uD\dlF261\u5\d _ .TE .nr PS 9 .RT .ad r \fBTableau B\(hy2/X.136 [T16.136], p.\fR .sp 1P .RT .ad b .RT .sp 1P .LP B.3 \fINotes on key assumptions, results and implications\fR .sp 9p .RT .PP For further study. .RT .LP .rs .sp 21P .ad r BLANC .ad b .RT .sp 2P .LP \fBMONTAGE:\ \ \fR Rec. X.137 sur le reste de cette page .sp 1P .RT .LP .bp