home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Internet Info 1997 December
/
Internet_Info_CD-ROM_Walnut_Creek_December_1997.iso
/
ietf
/
ptopomib
/
ptopomib-minutes-97aug.txt
< prev
Wrap
Text File
|
1997-10-10
|
15KB
|
399 lines
OPS Area
PTOPOMIB WG Meeting Minutes
39th IETF
Munich, Germany
August 13 & 14, 1997
Minutes by Andy Bierman
Summary
-------
The PTOPOMIB WG met to advance progress on all I-Ds under development:
- PTOPO MIB
- PTOPO Discovery Protocol and MIB (PDP)
- Entity MIB Extensions for Persistent Identifiers
Review Material
---------------
(1) Physical Topology MIB
<draft-ietf-ptopomib-mib-00.txt>
(2) Physical Topology Discovery Protocol and MIB
<draft-ietf-ptopomib-pdp-00.txt>
(3) Entity MIB Extensions for Persistent Component Identification
<draft-bierman-entmib-compid-00.txt>
Agenda
------
Wednesday:
Agenda Bashing
Issues - PTOPO MIB I-D
Issues - PDP I-D
Thursday:
Presentation: Physical Topology with standard MIBs
(Nick Dawes - Loran)
Issues - Entity MIB I-D
Wrap-up
Minutes
-------
This section contains a somewhat temporal account of WG issues and
discussions during the two meetings.
1) PTOPO MIB
1.1) Replacement of ptopoConnEntries
The current draft specifies that only one type of component
identifier should be present for each connection endpoint.
Entries with higher values of the ptopoChassisIdType or
PtopoPortIdType INDEX component are supposed to be replaced
by identifier types with lower values (if and when such
information is learned by the PTOPO agent).
The WG decided that it is desirable to allow an agent to
maintain multiple entries for the same connection, although
this is not mandatory. The MIB indexing will be modified
(see sec. 1.7) to allow an arbitrary number of ptopoConnTable
entries for each local port.
1.2) NULL Identifier Types
The WG discussed a proposal to allow NULL identifiers for
both chassis and port components. After much discussion
the WG decided there were enough identifier types to allow a
PTOPO agent to identify each component with a non-NULL string.
In the general sense, a NULL identifier in the PTOPO MIB will
be reserved to indicate the "don't know" value.
1.3) Repeater Backplanes
Since repeaters must transmit PDP advertisements out all ports
attached to the same shared backplane, it is not possible to
identify a single port in the PDP Port ID TLV and ptopoPortId
MIB object.
The WG decided that entPhysicalEntries with an associated
entPhysicalClass value of 'backplane(4)' should be supported in
the PTOPO MIB and PDP. Repeaters which generate PDP
advertisements should use the appropriate backplane
entPhysicalEntry in the PortID TLV, if the Entity MIB is implemented.
1.4) PTOPO Traps
A default value of 'false(2)' will be specified for the
ptopoConfigTrapsEnabled scalar object. Text will also be added
to the security section regarding this issue.
1.5) SNMPv3 Alignment
The ptopoConnNetAddr object specifies only a network address of an
SNMP agent containing information associated with the remote endpoint.
The WG discussed the need for additional MIB objects to support SNMPv3
command-initiator functions in a secure environment.
These objects (e.g., engineID, contextName, securityModel, securityName)
will not be added in the next draft of the MIB. The current table does
not include community strings either. It is considered a security risk
and burden to add this support to the PTOPO MIB and PDP.
1.6) ptopoChassisId and ptopoPortId TCs
The TCs which define PTOPO chassis and port identifier strings will have
a max length of 32 octets, instead of 48 octets.
1.7) ptopoConnTable INDEX
In order to allow an agent some flexibility to maintain multiple endpoint
entries (more than 2) per connection, the indexing of the ptopoConnTable
will be redone. The table will now be indexed by 4 integers:
INDEX { ptopoTimeMark, ptopoConnLocalChassis,
ptopoConnLocalPort, ptopoConnIndex }
*** NEW WG ISSUE ***
Local Index Uniqueness Requirements:
No Index Reuse:
The simple arbitrary local integer (ptopoConnIndex) must be unique
for a given { chassis, port } pair. An agent must insure that
each index is used at most one time per port between agent reboots.
New index values must be used each time an entry is added to the table.
This restricts the number of significant topology config changes
to 4 billion per port, but it makes the MIB easier to use for an NMS.
Allow Index Reuse:
Some of the same usage and limitations as the first approach, except
that an agent is permitted (required?) to reuse a ptopoConnIndex value
for an entry that has been aged out and then subsequently re-learned.
The new entry must have the exact same semantics as the previous one
at the time it was aged out.
The WG must decide which approach to use so the Editor can write the text.
1.8) Last Verification Timestamp
A MIB object will be added to the ptopoConnEntry to indicate
the value of sysUpTime when the indicated entry was last
verified by the PTOPO agent (e.g., by receipt of a PDP msg).
1.9) Max Desired Hold Time
A scalar MIB object will be added to the ptopoConfig group
called 'ptopoMaxDesiredConnHoldTime', which will specify
the number of seconds a PTOPO agent should maintain connection
information in the absence of entry verification.
If an entry exceeds the configured threshold, i.e.,
cur_sysUpTime > (ptopoConnLastVerifyTime + (ptopoMaxDesiredHoldTime * 100))
then the PTOPO agent is expected to remove the entry. Note that an
agent is permitted to remove entries using periodic garbage collection,
so entries may not be deleted at the instant the age-out threshold
is exceeded.
1.10) Statically Configured ptopoConnEntries
Entries which are statically configured by an NMS are not subject
to the garbage collection described in previous sections.
Static entries are indicated by a new read-create MIB object in
the ptopoConnEntry called 'ptopoConnIsStatic', with SYNTAX TruthValue.
There isn't any one appropriate DEFVAL, but entries with an
associated ptopoConnDiscAlgorithm value of { 0 0 } (conf by NMS)
are most likely static entries.
1.11) New MIB Counters
Three new scalar counters will be added to the ptopoConfig group to
help an NMS identify PTOPO agent activity regarding garbage collection
and topology changes. The ptopoConnTabInserts and ptopoConnTabDeletes
counters will remain exactly as defined.
A 'drop count' object will be added called ptopoConnTabDrops.
This counter indicates the number of times a PTOPO agent detected a
situation in which information would have been added to
the ptopoConnTable, but wasn't because of insufficient resources.
A 'replace count' object will be added called ptopoConnTabReplaces.
This counter indicates the number of times remote endpoint information
of one identifier type has been replaced by information of
a different identifier type.
An 'age-out count' object will be added called ptopoConnTabAgeOuts.
This counter indicates the number of times an entry has been removed
because the max desired hold time for the entry was exceeded.
The ptopoConnConfigChange notification will be updated to include
the new counters.
1.12) ptopoConnNetAddr Name Change
This object specifies a network address of an SNMP agent that contains
information associated with a remote connection endpoint. The object
will be renamed to ptopoConnAgentNetAddr.
1.13) ptopoConnChassis1,2 and ptopoConnPort1,2 Name Change
The chassis and port identifier objects will be renamed to differentiate
between 'local' and 'remote' connection endpoints. E.g., ptopoConnChassis1
will become ptopoConnLocalChassis and ptopoConnChassis2 will become
ptopoConnRemoteChassis.
1.14) TLV Support in the ptopoConnTable
The WG discussed a proposal to add MIB objects to support vendor-specific
MIB placeholders in the PTOPO MIB. The WG decided that even though
PDP supports vendor-specific TLVs, the standard PTOPO MIB should not.
Such objects should be defined in a vendor's proprietary MIBs.
1.15) Topology Server MIB
The WG recognizes that an aggregation of information gathered from
potentially many PTOPO agents are needed for applications to present
complete physical topology maps, and that future MIB work,
possibly in the DISMAN WG, is needed to effectively deploy
a standard topology server MIB.
1.16) NVRAM Requirements
The MIB will be updated to require agents which support NV storage
of static ptopoConnTable entries to recreate the exact semantics
upon agent restart, except that none of the same index values
have to be reused.
2) PDP
2.1) Multiple Sources for ChassisID and PortID TLVs
Since the PTOPO MIB supports multiple identifier types, the
chassis and port identifiers used in PDP TLVs should also
support this feature. The ability to associate an identifier
with any MIB object is already supported. It is an implementation
specific matter as to how co-located PDP and PTOPO agents
convert these OIDs to PTOPO identifier types.
2.2) PDP Forwarding
Text will be added to the document further clarifying the
differences between 'PDP forwarding' and 'PDP non-forwarding'
types of PTOPO implementations. An agent may contain entries
representing ports not directly attached to the indicated
local chassis. An agent may wish to limit and even suppress
such entries, but this is not required.
2.3) Selection of MAC address
In the event a MAC address is used in a chassis (or even a port)
identifier TLV, the agent may have to make an implementation-specific
decision as to which address to choose if more than one is available.
2.4) Elements of Procedure Appendix
An appendix will be added to the document to demonstrate how
a PTOPO agent may use info from a co-located PDP agent to
populate the PTOPO MIB.
2.5) PDP Checksum
The checksum in each PDP message operates exactly as a UDP checksum.
The PDP message is encoded with a checksum value of zero. Then the
ones-complement of the ones-complement addition of the encoded message
is rewritten to the checksum location in the message.
The checksum is optional, and the value zero is reserved to indicate
the checksum is not in use.
2.6) ASN.1 Encoding vs. TLV Encoding
The WG discussed a proposal to use ASN.1 and a pseudo-MIB to describe
the PDP message format and encoding. Proponents think it is overall
simpler to encode the PDP message payload as a varbind list
in ASN.1/BER encoding, rather than a list of TLVs in network byte order.
The WG held a straw vote on which encoding to use:
ASN.1 -- 6
TLV -- 3
don't care -- approx. 35
The WG decided to use ASN.1 encoding in PDP and describe the PDP
parameters using a pseudo-MIB (not actually implemented by any SNMP
agent), based on this straw poll.
Since the document editor doesn't know how to write a MIB object
for a UDP-style checksum (along with requisite encoding hacks),
the PDP header and checksum need to be outside the portion of
the PDU encoded as an ASN.1/BER varbind list. Only the encoding of
the TLV portion of the PDU will can be affected by this change.
2.7) Management Address TLV
A proposal was made to make the mandatory management address
TLV an optional TLV. As a compromise, the TLV will remain
mandatory, but an empty string may be used for the "don't know"
value.
2.8) Default Timer Values
A proposal was made to change the DEFVAL for the PDP transmission
interval from 60 to 10 seconds. Some wanted to raise the DEFVAL even
higher than 60 seconds. The WG agreed to leave the DEFVAL at 60,
noting that an agent may be configure any legal DEFVAL in implementation.
2.9) Replace pdpMessageTxHoldTime
This object will be changed to indicate a multiplier value, used with
the pdpMessageTxInterval object to derive the same semantics, except
that illegal configurations will be avoided. The DEFVAL will be 3,
and the same index range will be maintained.
2.10) Authenticated Discovery Protocol
A proposal was made to support the ability to encode the PDP messages
as auth/no-priv or auth/priv SNMPv3 PDUs. This was rejected due to
increased complexity and the inability to do administration such as
discovery protocol.
2.11) pdpTxSuppressTable Changes
The table used to suppress PDP message transmission will be generalized
to suppress PDP transmission and processing of any PDP messages
received on a particular port. The table will be renamed
to 'pdpSuppressTable'.
2.12) PDP MAC Group Address
An appropriate group address must be chosen for PDP messages.
Keith McCloghrie volunteered to act as WG liaison to the IEEE
802.1 sub-working group and pursue the matter further.
The IETF may have a group address allocated which can be used
for PDP. This option will be pursued by the WG as well.
3) Physical Topology Discovery using Standard MIBs
Nick Dawes of Loran reported on implementation experience
regarding the discovery of physical topology using MIBs
commonly found on various network devices.
The important features of this SW include:
- use of probability-based weighting factor to help resolve data
conflicts from multiple sources, which can occur in the normal
operation of the network.
- designed for minimal impact on network traffic; polling gated
by max-desired-bandwidth (as a percentage of bandwidth); only
minimum set of required MIB objects actually retrieved from
each agent.
4) Entity MIB Requirements
4.1) entPhysicalAlias object
The max length of this string will be 32 octets instead of 48,
to match other changes made in the PTOPO MIB.
The string will be based on the UTF-8 based SnmpAdminString,
defined by the SNMPv3 WG.
4.2) Entity MIB Document Advancement
The Entity MIB WG needs to do extensive work to support SNMPv3
and investigate RFC 2037 implementation experience.
Therefore, the PTOPOMIB WG wants the ENTMIB WG to phase
the upcoming controversial work, and publish a simple
augmentation to RFC 2037, containing just the entPhysicalAlias
extension defined in draft-bierman-entmib-compid-00.txt.
The ENTMIB WG Chair will discuss re-activation with the
Area Director, and report back to the PTOPOMIB WG.
Meeting Resolutions
-------------------
A) Document Advancement
New versions of the PTOPO MIB and PDP I-Ds will be published
by 17sep97, which will then be discussed on the mailing list.
The Entity MIB WG will hopefully be restarted in time to hold a first
meeting at the next IETF in December.
The group MAC address issues will be discussed on the mailing list
and hopefully resolved at the next IETF.
B) Interim Meeting
No interim meeting has been planned at this time. A charter extension
has been requested instead. Hopefully the WG can complete the currect
charter soon after the December IETF meeting.
--------------64865A6F589F43D0524F8516--