home *** CD-ROM | disk | FTP | other *** search
-
-
- CURRENT_MEETING_REPORT_
-
-
- Reported by Phill Gross/CNRI
-
- NJM Minutes
-
- Agenda
-
-
- o The role of the NJM WG
- o Request for statistics from researchers
- o Maps
- o Operations Area
-
-
- The role of the NJM WG.
-
- Historically the National Nets have been represented. Matt Mathis wants
- to hear national net information at NJM meetings. For example, SNMP
- Session names, and information on T3 migration. However, we agreed this
- should be more than just an NSFnet group.
-
- Although there is some overlap with TEWG, TEWG concentrates on topology
- while NJM concentrates on management and monitoring.
-
- Request for statistics by researchers
-
- Phill Gross has received several requests for stats from bona fide
- researchers. The issue is to get real data for models and simulations,
- and to do performance analysis. Gross asked how many regional networks
- kept stats. About 10, the majority of those present, kept stas.
- Mathis: PSC collects about 20 mbyte per month. Most goes to tape. PSC
- tends to get their needs met without requiring much analysis.
-
- Guy Almes listed 3 kinds of stats: 1) Reliability, 2) Character of
- usage (i.e., NNstat, end points pairs, protocol types), and 3)
- Performance, congestion.
-
- Finkelson: Most analysis programs are local with different formats. He
- uses NYSER package to collect SNMP data and uses graduate students for
- data reduction.
-
- That seemed typical. Data tends to be regional specific, and format is
- special, often reduced. Also, commercial clients of the nets may object
- to net management data being freely given out to researchers.
-
- Almes: ``Character of usage'' data is sensitive to some users.
- Perhaps, IRTF could characterize kinds of data they want and we could
- work toward provding it. Real data is best. Using real data is good
- for us too.
-
- 1
-
-
-
-
-
-
- Dan Wiverhan (OSU): OSU has tool that monitors multiple ethernets; runs
- on PC; tells protocols and end-points; Based on KA9Q; Supports SNMP.
-
- Long discussion ensued on various types of date collection. Gross: It
- either MIB or NNstat-based.
-
- Apparently, some vendors are reluctant because makes their boxes slower.
- Perhaps a smaller set of common stats would meet better acceptance from
- vendors. We tried to draw up a list of ``low impact, high yield''
- metrics. Perhaps we should develop a subset of MIB that everyone should
- archive in standard format, with standardized time granularity.
-
- Maps
-
- Questions: Should USWG catalog all on-line MAP sites? Should format be
- standardized (e.g., Dated to show currency, All line speeds shown)? Who
- is working on mapping techniques (MERIT and Bellcore)?
-
- Are maps really used to debug a problem to another site? They are often
- too far out of date. Many folks said they used maps to debug. Also
- useful in topology planning. Counter by Mathis: Maps often don't show
- interesting (i.e., surprising) links.
-
- Ted Brunner (Bellcore) is working on auto map generation. [He gave a
- demo that afternoon.] Uses MIB. For better display, will probably need
- more MIB objects than currently available. He has extended MIB to hold
- extra information. This became an interesting topic. Gross was asked
- to make sure this was on the agenda for the next meeting.
-
- Other efforts? Gross and Enger mentioned the Contel ``net-feeb''
- program. People are interested. Will contel make available?
-
- Bottom line of this topic - Maps are potentially very useful, but would
- be much better with better methods (e.g., common formats, auto
- generation, up-to-date).
-
- Operations Area
-
- Gross: This may not be in NJM Charter, but this group can give
- important feed-back/advice. Should IETF Operations Area be pro-active
- or re-active? Formal or informal? Liason only? Should IETF propose a
- set of guidelines for Internet operations? What other groups should be
- involved? Should we reach to local managers directly, or through
- Farnet? Should we identify one Area Director or ``Board of Directors''?
-
- Almes: More operations folks need to attend the standards sessions.
- Gross: Differences between ANSI and IETF standards setting process.
- More user and operations input at IETF.
-
- Bottom line -- An active operations area is important. Interaction with
- protocol development is very important. See the operations area report
-
- 2
-
-
-
-
-
-
- for more detailed description of the group consensus about the
- operations area. The IETF Chair gives his thanks to the NJM Working
- Group for helping to formulate the direction for the IETF Operations
- Area.
-
-
-
- 3
-