home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Internet Standards
/
CD2.mdf
/
ccitt
/
1992
/
e
/
e862.asc
< prev
next >
Wrap
Text File
|
1991-12-30
|
35KB
|
645 lines
All drawings appearing in this Recommendation have been done in Autocad.
Recommendation E.862
DEPENDABILITY PLANNING OF TELECOMMUNICATION NETWORKS
Introduction
This Recommendation is concerned with models and methods for dependability
planning, operation and maintenance of telecommunication networks, and the
application of these methods to the various services in the international
network.
The CCITT,
considering
(a) that economy is often an important aspect of dependability planning;
(b) that the ability of achieving a certain level of dependability differs
between network providers;
(c) that network providers often operate in a competitive environment;
(d) that Recommendations E.845, E.850 and E.855 establish objectives for
serveability performance;
(e) that objectives for dependability performance are deducible from
Recommendations Q.504, Q.514, and X.134 to X.140;
(f) that these objectives have been established in an intuitive manner
rather than based on analysis of user needs;
(g) that there exists no unambiguous way of implementing these objectives
in planning;
(h) that there is a need of establishing a method for dimensioning and
allocating dependability in the telecommunication network;
(i) that terms and definitions relevant to concepts used for dependability
may be found in Recommendation E.800,
recommends
that the procedures defined in this Recommendation shall be used by
Administrations to plan, design, operate and maintain their networks.
1 General
Dependability planning may be accomplished by using essentially two
different methods.
Intuitive method
The level of dependability is determined by making a synthesis of
objectives and procedures presently used. It is a pragmatic method in absence of
an analytical method or in the case when necessary data for a thorough analysis
is not available.
This method reflects the present status, but is inconsistent in achieving
what Administrations actually want to attain: the most economic level of
dependability taking into account customer needs and inconvenience.
Fascicle II.3 - Rec. E.862 PAGE1
Analytical method
The analytical method is based on principles defining the object of
dependability planning. The principles are realized through a quantitative model.
The level of dependability is deduced by applying the model, taking into account
all relevant factors in each planning case.
- Basic principle: The main object of dependability planning is to find a
balance between the customers' needs for dependability and their demand
for low costs.
- Model: Fault consequences are expressed in terms of money and are
included as additional cost factors in planning and cost-optimization.
The cost factor reflects the customers' experience of faults in the
network, quantified in terms of money, as well as the Administration's
costs for lost traffic revenue and corrective maintenance.
- Application: The Administration is provided with a method to integrate
dependability as a natural part of planning, taking local information
from the actual planning case into account. This method enables the
preparation of simplified planning rules.
The application of the analytical method gives, economically, the
best-balanced level of dependability, seen from the customer's point of view.
This reduces the risk of customer complaints and loss of business to competitors
as well as the risk of unnecessary investments. It is therefore considered to be
the best general way of planning dependability for the Administration, as well as
for the customers.
Recommendations for operational dependability objectives are needed in
order to discover impairments and to check and compare dependability performance
in the national and international network. Experience from the application of the
analytical method may give reason to revise existing Recommendations.
2 Generic measures for dependability planning
The dependability is described by measures defining the availability
performance, the reliability performance and the maintainability performance of
the network and its constituent parts as well as the maintenance support
performance (for the maintenance of the network). The recommended measures are:
Availability performance
Mean accumulated down time
Reliability performance
Mean failure intensity
Maintainability performance
Mean undetected fault time
Mean time to restoration
Mean active repair time
Maintenance support performance
Mean administrative delay
Mean logistic delay
Note - The definitions of these measures are given in Recommendation E.800
and Supplement No. 6.
3 Characteristics of network faults
The faults occurring in the telecommunication network are characterized
mainly by their impact on the service provided by the network, i.e. by the
traffic disturbance they cause. Important measures determining the traffic
disturbance due to a fault are:
Duration of the fault (mean down time), T in hours (h)
Mean traffic intensity affected by the fault, A in Erlangs (E)
Mean probability of congestion during the fault, P
The seriousness of a fault also depends on how the customers experience
the fault, and on the Administration's loss of revenue. In order to express this
fact, the value of a unit of traffic volume (Eh) disturbed by the fault is
quantified in economic terms.
Measure: the economic valuation of affected traffic volume is c (monetary
units per Eh).
A number of factors may influence this variable such as:
- the category of customers and services affected,
- the degree of congestion or transmission disturbance during the fault,
- the duration of the fault,
- the accessibility to alternative communication means for the affected
customers,
PAGE12 Fascicle II.3 - Rec. E.862
- time of day, week or year when the fault is in effect,
- how often faults have occurred in the past.
Additionally, the Administration's costs for corrective maintenance also
contribute to the assessment of fault consequences.
Measure: the maintenance cost per fault is cm (monetary units per fault).
4 Planning for economic optimum
4.1 Economic dimensioning and allocation method
Mathematically expressed, the main principle of dependability planning is
to find actions that minimize the total cost of the network:
eq min \b\bc\{ ( \a\co(\s\do15(CI + Cm . d + Ct . d + . . . )))
where
CI is the investment costs to achieve a certain degree of dependability,
Cm is the expected annual costs for corrective maintenance,
Ct is the expected annual traffic disturbance costs,
d is the discount factor for calculating the present value of the annual
cost over the lifetime of the investment.
Ct reflects the annoyance caused by faults and should be regarded as the
basic service parameter which dimensions and allocates dependability in the
network under given conditions.
An action is optimal if the following two conditions are met:
1) The benefit of the action (e.g. lower traffic disturbance cost) is
larger than the cost, i.e. the action is profitable.
2) The action is the best in the sense that the ratio benefit/cost is
maximal. There are no alternative actions that give a higher profit.
The method points out a profit seen from the customer's point of view,
i.e. the actions are not necessarily profitable for the Administration in the
short run. Rates and charges might therefore have to be increased to finance the
actions. However, satisfying the customer's needs is recommended as the generally
most profitable policy for the Administration in the long run.
This method is applicable for planning all parts of the national and
international network and for dimensioning the dependability of network
components and the level of the maintenance support. It may be used in short term
planning as well as in long term optimization and strategic planning.
The method does not become out of date with technological advances,
changes in cost structure etc. Dependability is converted to one clear-cut
measure (money) which makes it easier to evaluate actions to promote
dependability and to compare and choose between different alternatives.
4.2 A model for traffic disturbance costs
The annual traffic disturbance cost is obtained by multiplying the
disturbed traffic volume (lost, delayed or affected by transmission impairments)
by the monetary valuation of disturbed traffic volume c and the mean failure
intensity z which gives:
Ct = PAzTc
where
T is the duration of the state of increased congestion or transmission
disturbance due to the fault, mainly the down time. Congestion due to
traffic overload after the fault has been repaired might however also
have to be included.
A is the intensity of offered traffic.
P is the portion of the offered traffic volume over the time T, delayed
or lost.
z is the mean failure intensity.
c is the monetary valuation of disturbed traffic volume. c may be
dependent on any number of factors, i.e. c = c(P, T, A, . . .).
Assuming traffic variations, A(t), and consequently variations of
congestion, P[A(t)] = P(t), then A and P are calculated as follows:
eq P = \f(\i(0,T, ) A(t) P(t) dt ,\i(0,T, ) A(t) dt)\D\FO30() A =
\f( 1, T)\D\FO10()\i(0,T, ) A(t) dt
Normally it is not possible to predict the instant of time when a failure
will occur. In this case A is a long time average incorporating yearly variations
and long time trends. P is calculated by using an average traffic profile.
Fascicle II.3 - Rec. E.862 PAGE1
Recommendations E.506, E.510 and E.520 to E.523 deal with methods for traffic
calculations.
4.3 Economic assessment of disturbed traffic volume, c
The factor c reflects the level of ambition of an Administration or
Operating Company in dependability planning. A high valuation of c will give a
high level of dependability and vice versa. The values used by an Administration
are related to the society's dependence on telecommunications which in turn might
be dependent on standard of living, national economy, price level, etc. The
establishment of c on the national level is therefore a national matter.
However, it is recommended that c should reflect the combined experience
of the Administration and the customer, i.e. it should consist of:
1) the Administration's loss of revenue due to traffic not recurring after
the fault,
2) an assessment of the average customer's economic loss due to a unit of
traffic volume (Eh) being affected by a fault,
3) a symbolic price tag reflecting the annoyance experienced by the
average customer.
The sum of 2) and 3) should reflect the price the average customer is
willing to pay to avoid one Erlang-hour of offered traffic, delayed or lost due
to a fault. The result will then be a level of dependability the customers are
satisfied with and prepared to pay for.
Administrations are recommended to make their own investigations among
their customers in order to determine the values to be used for planning. Annex B
gives an example of such an investigation.
If this is not possible, rough estimates may be obtained from information
about actions taken in the present network. The cost of an action is compared to
the amount of traffic it saves. Actions that intuitively are regarded as
reasonable give a lower limit of c, actions that obviously are unreasonable give
an upper limit. The values derived this way are then used in optimization under
the assumption that they are valid also for planning the future network.
If c is not possible to estimate at all, the method may still be used to
find an optimum allocation of a given amount of resources. The level of
dependability attained in this case does however not necessarily satisfy the
customers.
4.4 Planning procedure
Traffic disturbance costs are included as additional cost-factors in
economical calculations for planning, thus integrating dependability as a natural
part of planning.
The procedure of dependability planning is performed in four steps:
Step 1: Plan a network attaining functional and capacity requirements.
The starting point is a network planned and dimensioned in order to comply
with the functional and capacity requirements, but without special
consideration of dependability (zero-alternative). The second step is to
identify what changes may be necessary to promote dependability.
Step 2: Search for actions to promote dependability.
There is a need for actions to promote dependability if traffic
disturbance costs are high or if the actions can be taken at a low cost. A
non-exhaustive list from which actions could be identified is given below:
- Protection of equipment in order to prevent failures
- Choice of reliable and maintainable equipment
- Modernization and reinvestment of worn out equipment
- Redundancy
- Overdimensioning
- Increase in maintenance support
- Network management actions to reduce fault effects.
Step 3: Analyse the actions.
Express improvements in terms of changes in traffic disturbance and
maintenance costs (DCt + DCm) for each action. It is only necessary to
calculate costs that differ between the alternatives. Annex A gives
examples of dependability models for network design, maintenance support
planning and for determining requirements for network components.
Compare DCt + DCm to the increased investment cost (DCI) for each action,
e.g. by the present value method.
Choose the best set of actions, i.e. which gives the lowest total cost.
Step 4: Check that minimum requirements are complied with.
PAGE12 Fascicle II.3 - Rec. E.862
A minimum service level may be stipulated by governmental regulations, by
CCITT Recommendations, for commercial or for other reasons. The
establishment of any minimum requirements on the national level is a
national matter. For planning of the international network the
Administration is recommended to check if dependability objectives
deducible from existing CCITT Recommendations are met. If not, the reasons
for non-compliance should be examined more closely. If it is justified,
the level of dependability should be adjusted.
4.4.1 Numerical example based on the above
Step 1: Network planned without special consideration of dependability.
The network studied is the trunk between two exchanges.
Figure
Step 2: Search for actions to promote dependability.
The action considered is to introduce a physically redundant cable. It is
assumed to be dimensioned to carry the whole traffic load, i.e. a single
failure will not disturb the traffic.
Figure
Step 3: Analyse the action.
Assumptions
Failure intensity z = 0.1 failures/year
Mean down time T = 24 h
Mean offered traffic A = 100 E
Congestion P = 1 (without redundancy)
P = 0 (with redundancy)
Monetary valuation of disturbed traffic volume c = 400 monetary
units/Eh
Discount factor
(lifetime 25 years, interest 5% per year) d = 14
Maintenance cost per failure cm = 1000 monetary
units/failure
Cost of redundant cable CI = 400 000 monetary units
Calculations
Traffic disturbance costs for network without redundancy:
Ct = P . A . z . T . c = (1) (100) (0.1) (24) (400) = 96 000 per
year
Present value Ctd = (96 000)(14) = 1 344 000
Traffic disturbance costs for network with redundancy (the possibility of
simultaneous faults is negligible):
Ct = 0
Change in traffic disturbance costs:
DCtd = 0 - 1 344 000 = -1 344 000
Maintenance costs without redundancy:
Cm = zcm = (0.1)(1000) = 100 per year
Present value Cmd = (100)(14) = 1400
Maintenance costs with redundancy:
Cm = 2zcm = (2)(0.1)(1000) = 200 per year
Present value Cmd = (200)(14) = 2800
Change in maintenance costs:
DCmd = 2800 - 1400 = 1400
Cost reduction:
DCtd + DCmd = -1 344 000 + 1400 = -1 342 600
Change in total cost:
DCI + DCmd + DCtd = 400 000 - 1 342 600 = -942 600
Conclusion
Since DCI + DCmd + DCtd < 0, the action is profitable. Whether or not it
is optimal depends on whether there are alternative actions that are more
profitable.
Step 4: Check minimum requirements
Any additional actions to meet governmental requirements (for defence
Fascicle II.3 - Rec. E.862 PAGE1
reasons, emergency, etc.) should be taken.
5 Applications to the international network
5.1 Value of c for international traffic (for further study)
In order to dimension and allocate dependability to different parts of the
international network a uniform way of evaluating affected traffic should be
established. It is recommended that the following values (ci) be used as a guide
in the planning of the international network
ci = xi SDR : s/Eh (values to be determined)
The values refer to a particular reference year. Price increase due to
inflation, society's increasing dependence on telecommunication etc., should be
taken into account.
5.2 Planning recommendations (for further study)
When values of c have been established, it is possible to make economic
dependability analyses of the international network. These studies may be done in
a similar manner and using partly the same data as for cost studies of charging
and accounting.
The object of the studies is to arrive at planning recommendations, e.g.
for the amount of redundancy, maintenance support, etc., in different parts of
the international network.
5.3 Operational objectives for dependability (for further study)
The result of the economical dependability analysis of the international
network is presented in terms of reliability, maintainability and maintenance
support performances of different parts of the network. This will help
Administrations monitoring and checking their networks to discover impairments,
misplanning, etc.
ANNEX A
(to Recommendation E.862)
Simplified models for dependability planning
A.1 General
The object of this annex is to show simple examples of how different
models of dependability may be used to calculate traffic disturbance costs and
how the calculations can be used in planning. A list of actions is given in S
4.4. The applications may be divided into:
- Network planning (SS A.2 and A.3)
- Dimensioning dependability of network components (S A.4)
- Maintenance support planning (S A.5).
A.2 Example: Redundancy
The traffic disturbance cost of a redundancy consisting of two independent
items as shown in Figure A-1/E.862 is:
Ct = P1z1T1Ac(P1) + P2z2T2Ac(P2) + z1z2T1T2Ac(1)/8760
where
P1 is the average congestion when item 1 is faulty,
P2 is the average congestion when item 2 is faulty.
Figure A-1/E.862
A simple case is when the two items are identical and each can carry the
whole traffic load (see Figure A-2/E.862), then:
Ct = z2T2Ac(1)/8760.
Figure A-2/E.862
By installing a redundant item, the traffic disturbance costs are reduced
by
DCt = zTAc(1) - z2T2Ac(1)/8760.
The second term is often negligible, thus DCt may be approximated by DCt =
zTAc(1).
A.3 Example: Optimal dimensioning for diversified routes
The problem is to determine the optimal number of channels, N1 and N2
respectively, for which the two redundant routes should be dimensioned, see
Figure A-3/E.862.
Figure A-3/E.862
Denote CN to be the cost per channel. The optimal allocation of channels each way
is found by solving
eq \o\ac(min,\s\do10(N1\l(,)N2))\b\bc\{ ( \a\co(\s\do15((N1 . CN1 + N2 .
CN2) + (P1 . A . z1 . T1 . C(P1) + P2 . A . z2 . T2 . C(P2)) d)))
PAGE12 Fascicle II.3 - Rec. E.862
This implies an overdimensioning in the fault free condition. The benefit
of this is not included in this formula. The effect of simultaneous faults does
not influence the optimization.
Fascicle II.3 - Rec. E.862 PAGE1
A.4 Example: Optimal testing time
Assume that the failure intensity z(t) after a certain operation time (t)
is given by
z(t) = z0 + ze-bt
where
z0 + z is the failure intensity at t = 0,
z0 is the constant failure intensity after the early failure period,
b is the factor determining the decrease in failure intensity during the
early failure period.
By testing, faults may be corrected before causing traffic disturbance and
maintenance costs. Assume that:
cm + ATc are the maintenance and traffic disturbance costs per fault,
C is the cost per year of testing.
The optimal testing time (t`) is found by solving
eq \o\ac(min,\s\do10(t)) \b\bc\{ ( \a\co(tC +\f(z,b) e-bt(cm +
ATc)))
where
eq \f(z,b) e-bt is the additional number of faults occurring in operation
as a function of the testing time.
Optimal test time: t` = eq \f( 1, b) ln eq \f( z(cm + ATc), C) .
A.5 Example: Optimal number of maintenance units
Mean delay w(N) as a function of the number of maintenance men (N) may in
some cases be mathematically expressed by using queuing theory. The simplest case
is if the times between failures and repair times are exponentially distributed
(an M/M/N queue model). w(N) is obtained by calculating:
w(N) = eq \b\bc\[ (\f( (z/m)N . m, (N - 1) ! (Nm - z)2)) / eq
\b\bc\[ ( \i\su(N-1,k=0, )\d\fo4()\f( 1, k!)\b\bc\( (\f( z, m))\s\up6(k) +\f( 1,
N!)\b\bc\( (\f(z,m))\s\up6(N)\b\bc\( (\f(N,Nm - z)))
where
N is the number of maintenance units,
z is the intensity of failures,
w(N) is the mean delay as a function of N,
A is the affected traffic intensity,
c is the valuation of affected traffic volume,
m is the repair rate.
The model may be refined by taking into account classes of priority. It is
also possible to let faults of a higher priority interrupt assignments with a
lower priority.
If CN is the annual cost per maintenance unit, the optimal number of
maintenance units is obtained by solving:
eq \o\ac(min,\s\do10(N))\d\fo5()\b\bc\{ ( \a\co(\s\do15((NCN +
zw(N)Ac))))
ANNEX B
(to Recommendation E.862)
Example of an investigation to assess
the monetary valuation of disturbed traffic volume, c
B.1 The aim is to arrive at cost data to assess c. Different customer groups
and their monetary valuation of total and partial failures with respect to
typical traffic relations and different services is studied. Investigations are
carried out among residential and business customers based on the following
assumptions:
a) The customers are affected by telecommunication interruptions in mainly
two ways: in terms of annoyance and in terms of direct costs.
b) For residential customers, annoyance is likely to predominate. For
business customers, the direct cost may be important.
c) Both costs and annoyance increase by the duration of the interruptions
and the amount of traffic disturbed.
d) As a natural consequence of the great variations in dependence on
telecommunications there is a great variation of costs and annoyance.
e) Residential customers are not able to quantify their annoyance in
PAGE12 Fascicle II.3 - Rec. E.862
monetary terms. Faults on the home telephone mostly result in
irritation, and not in direct costs (except in the case of long-time
faults).
B.2 Complete faults
B.2.1 Business traffic
Companies chosen at random are asked to answer the following question:
"What is the estimated approximative cost of a total interruption of the
telephone or data service in connection with down times of 5 minutes, 1 hour, 4
hours, 8 hours, 24 hours and 3 days?"
Companies with experience of a specific fault are asked the question:
"What was the estimated cost of the fault just experienced?"
An estimate of the affected traffic intensity in connection with total
interruptions can be made on the basis of the number of exchange lines and the
number of data terminals for communication of each company, together with
information on how trunks are dimensioned and measurements on the calling
intensity of various customer classes.
On the basis of a stated cost, c is estimated according to the formula:
c = eq \f( (cost stated by the customer), (mean traffic intensity)
(down time))
Average values of c for telephony and data traffic are calculated for
different trades by means of a market profile (distribution of workplaces by
trade).
B.2.2 Residential customers
Group discussions on interruptions can be held in order to arrive at a
reasonable valuation. If there is little willingness to pay for increased
dependability a relatively low value of c is assigned.
B.3 Partial faults
A partial interruption of a traffic relation results in costs for the
customer mainly in the form of delays to commerce. By using a calculated hourly
salary this cost is estimated for business customers. On the basis of information
about the amount of business and household traffic, an average value of c for
traffic disturbed by partial faults is obtained.
B.4 Results
Table B-1/E.862 gives a few examples of figures derived by the Swedish
Administration. The figures have been used in various planning cases. The
Administration's loss of revenue is included in these figures. The cost figures
and exchange rate relate to 1986-01-01 [1 SEK (Swedish Krona) 0.1 USD (US
Dollar)].
TABLE B-1/E.862
Economic assessment of prevented communication (c)
Class of failure
Field of application Complete fault Partial fault
(P = 1) (P < 0.5)
Business customers with a large portion of data 1000 SEK/Eh 250 SEK/Eh
traffic
Used in the long distance network 400 SEK/Eh 100 SEK/Eh
Customers in a sparsely populated area. High cost 200 SEK/Eh 50 SEK/Eh
for alternative communication
An average value for areas with mostly 100 SEK/Eh 25 SEK/Eh
residential customers
Residential area where it is easy to reach 30 SEK/Eh 10 SEK/Eh
essential services. Low costs for alternative
communication
Fascicle II.3 - Rec. E.862 PAGE1