Meeting of the Operations and Management Area
IETF46, Washington DC, November 7-12 1999
Minutes reported by David Harrington
Working Group Status
- Distributed Management (DISMAN) WG has requested advancement of script and scheduling SNMP MIBs for adding script-support and scheduling support to mid-level/proxy managers (intelligent agents). The IESG is awaiting implementation experience reports.
- Physical Topology (PTOPO) WG has been inactive. It will be cancelled in two weeks, unless somebody objects.
- Authentication, Authorization, and Accounting (AAA) WG was re-chartered to focus on NASREQ, MobileIP, and RoamOps requirements, with an urgent need for results. Once requirements have been identified and the input has converged, the WG will be re-chartered for the next step.
- Network Access Server Requirements (NASREQ) WG has two documents going to last call. A third, the criteria document, is in the process of being wrapped.
- Roaming Operations WG is currently in shutdown mode, expected to be done by Mar00. Some remaining tasks have been transferred to the AAA WG.
- Routing Policy (RPS) WG has one draft moving to proposed standard
- SNMPv3 has three documents from SNMPv2 [rfc1905-1907] that are being updated before going to last call. They are expected to achieve full standard status by March. Deployment experience is being sought before the five SNMPv3 docs will be allowed to advance. [See the SNMPv3 meeting report below for more on this.]
- RMON MIB is finishing up its charter, and is identifying the focus of a new charter. The new charter will probably be focused on performance for application protocols, and mechanisms to make RMON easier to use, such as persistent address mapping, Diffserv extensions, and protocolDir optimizations.
- Agent Extensibility (AgentX) WG has two drafts about an SNMP master/subagent protocol advancing to proposed standard.
- ADSL MIB has one draft being published as RFC.
- Next Generation IP Transition (NGTRANS) WG has five documents on the way to the standards track. They describe IPv4/IPv6 transition mechanisms, gateway approaches, and tunneling.
- MBONE WG has a bunch of documents plodding along for PIM-SM, MBGP, GLOP, and MSDP. The WG is gathering and analyzing deployment reports, in preparation for asking for advancement.
- Benchmarking WG is chartered to define some standard metrics and methodologies for benchmarking. The have completed the RFC2467 Firewall benchmark metrics. ATM, Frame relay, LAN Switch, and Multicast documents are in progress.
- DNS Operations (DNSOPS) WG is behind their milestones slightly. The June/July items have been completed. The Root Server Requirements document is about to enter last call.
- GRIP has three documents in the works. The "expectations" document will be published as informational; two others will be updated by end of year.
- Traffic Engineering WG has recently moved from the Transport Area into the Operations and Management Area. It has been assigned various work items that had been started in other WGs.
- Y2K WG is in a holding pattern, waiting for Y2K to see if their preparation has been adequate.
MBONE deployment reports.
Dave Meyer [slides]
Many of the protocols have been deployed:
- Pim-sm > 4.5 years
- Mbgp 1.5 yrs
- MSDP SA-messages
- 6000+ network entities
- 200+ inter-domain groups
- 5-20 Mb
The WG is focusing on deployment problems. The biggest problem is how to debug. It is in the best shape it has ever been in, but there is still difficulty in debugging implementations and deployment. There are still some difficulties in inter-domain MSDP. There is a need for engineering tools to simplify debugging.
Dave Meyer [slides]
There is a BOF to decide what AnyCast encompasses and to narrow the scope of what AnyCast means.
There are a number of open issues about AnyCast. These include AnyCast address aggregation, how to specify AnyCast addresses for existing protocols, issue so fragmentation, and mapping to security associations.
Glen Zorn [slides]
Roaming Ops is currently in shutdown phase, which should be complete by Adelaide.
Milestones have been removed from the charter. AAA will probably define a protocol to satisfy RFC2477. Maybe an individual contribution on BCP for roaming will be done sometime.
Some RFCs have been published. More are coming on how to identify proxy chaining, and a set of radius attributes. Still waiting on ADIF and Phone Book.
David Perkins [slides]
A GR303 BOF is being held to gauge interest in an SNMP MIB to support Telcordia's GR303 specification for local exchange voice gateways to broadband access networks. A number of companies are doing work in this area. The spec calls for CMIP, but many customers want to use SNMP instead of CMIP.
Kevin Dubrai [slides]
The goals are to define metrics and terminology, the to define methodologies to get metrics.
The WG has completed the RFC2467 Firewall benchmark work.
ATM, Frame relay, LAN switch, and multicast documents are in progress.
They held their meeting the night before. There were reports on CAIRN, on policy and DNS structure, on key handling workshops, and on shared root server work. There were some new ideas proposed on CIDR delegation subclass C namespace modifications. DNS security tools and guidelines are needed so people can deal with DNS more easily. There was a discussion of what is technically required for DNS delegation, as compared to the political requirements. Some technical requirements should be documents, then other requirements can build on top of that.
DNSOPS is a bit behind on their milestones. They completed the June/July items. They are about to do Last Call on the Root Server Requirements document. Performance work will be delayed. It is hoped that the key handling work will be finished within the timeline.
There are three documents in the works. Expectations will be sent to the IESG for final handling as Informational or BCP. SSH-Add updated draft is expected by end of year; hopefully it will be the final doc. User-02 update is expected by end of year.
The WG will not meet in Adelaide unless something comes up.
Congratulations from Bert for lots of work completed in 39 days.
The AAA WG was re-chartered to focus on network access, due to an urgent need for results. The AAA documents incorporate by reference many documents from NASREQ, RoamingOps, MobileIP, and TIA. The requirements from the many groups have been brought together, and are still under discussion. The AAA will act as the summary group.
An initial policy/authorization architecture has been proposed.
The 39 day deadline got AAA well-focused, but a prevailing question is "What happens next?" Randy Bush said he thought the WG did amazingly well. However, asking other WGs for input is good, but joined requirements do not define an architecture, and the AAA WG is responsible for an overall architecture. It must yet be determined how we move to a coherent spec, and then where the work gets done to create protocols will be determined. Once the WG completes its current charter, it will need to be re-chartered, possibly into multiple areas.
There are three current drafts. Nasmodel draft is in WG last call. The radiusPract draft will go to WG last call following the meeting.
The Criteria document is wrapping up. It will serve as input to AAA, and needs some minor revisions. Barring significant problems, it should go to IETF last call by end of year.
The WG will be reviewing candidate AAA protocols from the standpoint of NASREQ. There may be no good protocols for NASREQ-AAA criteria.
The WG is a month or so behind on review and submission of documents. Once these are complete, we will be up for charter review.
Bob Fink [slides]
There are five documents on the way to standards track.
Transition Mechanisms is at IESG last call. The WG attempted to prune away multiple approaches implemented. It is unclear how the various mechanisms fit. The WG is waiting for implementation experience, and they will probably do interim meetings.
Stateless ICMP/translator documents discuss how to do stateless translation between IP and ICMP. This is used by other spec; it is not used alone. NAT-PT uses this, and they are trying to conclude IESG review.
The document on 6-4 clouds describes how to tunnel across the networks. This may be a savior, letting us have Ipv4 at the edges.
The DualStack transition avoids translators. It is too early to know how this will pan out.
There are some Informational documents in progress, to explain how to do transition.
The Bump in the sack and the SOCKS-Gateway approaches both insert bumps. They allow you to use Ipv4 apps that haven't been converted. An Ipv6 app means it must have been tested, and is not bound by 32-bit addresses. We must verify that all apps can work over Ipv6. Bumping helps vendors get around requirements.
Categorizing translators document is not part of translators, or standard. It is just information.
The Tunnel Broker doc describes an easy way to get tunnels into web/combination methods. A Canadian vendor reports having done 3000+ implementations. A European site has done some as well.
Regarding the 6Bone Routing Backbone - it is time to stop being a playground, and start pruning out people that shouldn't be in the 6Bone Backbone. Get rid of unreliable implementations.
A 6Bone Pre-qualification for address prefix allocation was discussed in Oslo. There has been a land rush for addresses anticipating future value, not for current use. There is an agreement for pre-qualifying requests from registries based on production address space.
A Guide to Introduction document will be a core document to explain how to transition.
There was discussion about which area Traffic Engineering belongs in. It ended up in Ops. There is one pending item due this month - a
framework. There are lots of hand-me-downs from other working groups that has been moved into Traffic Engineering.
Wes Hardaker [slides]
The most important item for tomorrow's meeting - we need deployment reports to prove to the community that snmpv3 helps to manage
networks. Implementers, please talk to your customers about reporting their deployment experiences in networks.
RFC 1905-1907 are being updated for advancement.
Andy Bierman [slides]
The WG is finishing up its charter.
SMON was recently finished. Some work has been done to extend to high-capacity RMON. A draft will move forward soon. All three
documents have been through WG last call, and will be going to the IESG.
There are two docs about RMON protocol identifiers. The protocol refernce will remain on the standards track. The protocol macros will be
moving to informational status.
RMON1 will advance to full standard. The WG last call is done. Implementations reports are on the web site. RMON is a $1billion industry, so
deployment is evident.
The WG is planning to ask for a new charter. The most important item is performance for application protocols. We also need features to make
RMON easier, such as persistent address mapping, Diffserv extensions, and protocolDir optimizations.
Configuration Management BOF
Luis Sanchez [slides]
This BOF is being called to discuss how network-wide configuration, especially for Policy, should be done.
About 45 days ago, the Area Directors called a closed meeting to discuss policy management within the context of network management. Several WG chairs and BOF chairs were invited to discuss terminology and use cases, etc., to reach a general consensus on what Policy covers. At the end of two days, two design teams were created to research 1) terminology, and 2) requirements.
The result of the design composed of Luis Sanchez, Jon Saperia, and Keith McCloghrie was a first cut at Configuration Management requirements, plus an evaluation of the SNMP/MIBs approach and the COPS-PR/PIBs approach.
The purpose of the BOF is to collect input on configuration management requirements, to discuss which protocols should be used, to discuss whether existing protocols should be modified or new protocols developed, and to provide feedback to the IESG so they can determine how efforts concerning configuration management should proceed.
The BOF will have presentations on the requirements, as currently understood, followed by short presentations about how the COPS-PR/PIBs approach meets the requirements, and how the SNMP/MIBs approach meets the requirements. That will be followed by discussion.
While this is a controversial topic, Bert expressed the hope that rather than controversy, he'd like to see contributions and results.
High-Capacity Data Types for SNMP
Andy Bierman [slides]
Andy described a proposal to resolve an outstanding problem of handling 64-bit data in SNMP. There is no way to define the needed mechanisms legally without forcing the already full-standard SMI to be reissued, and to wait the two years or so required to get back to full-standard status. New textual conventions that are "not quite legal" are proposed to provide a "snapshot" of a counter64, a difference of two counter64s, and a zero-based counter64. These are needed immediately to allow RMON MIB to advance.
The proposed solution will be published as a draft and will then start a four-week WG last call. This solution is a short-term focused solution. People are needed to draft a charter for a new WG to define a strategy for resolving this type of problem (extending SMI data types) consistently for the long-term.
Network Management Research Group (NMRG)
Juergen Schoenwalder [slides]
Juergen did a presentation of their research into network management extensions for SNMP.
For Bulk Transfer, they have researched an SNMP/TCP mapping, PDU compression, Table retrieval, and a MIB-based subtree retrieval mechanism. Much of their research has been documented in the Simple Times March 1999 issue. See www/simple-times.org
They have been researching a new SMI. There are still some problems with SMIv2, and they have experimented with some approaches to solve some "nasty things" in the current SMI. The SNMP community has expressed some interest in being able to extend the SMI with new data types, so they have been studying ways that could be done and remain backwards compatible. They also have been studying an ABNF-based syntax, to replace the cryptic ASN.1 syntax, while maintaining backwards compatibility through translator applications.
They have been researching adding Actions to SNMP. They have just started doing this research.
Service Management Research Group (SMRG)
Michael Eder [slides]
Michael did a presentation explaining what they did, with little technical information. They are chartered to work on the convergence of system and network management. They are looking into policy management to fulfill service management needs.
There will be no OpsArea meeting at the next IETF meeting. There will be one for the Summer meeting.
If there is a need to discuss new directions, etc., that can be done through the ops mailing list.
RMONMIB WG Status
HC-DATA Background (1/2)