Minutes of July 25, 1991 MJTS meeting

[Allan.C.Rubens]

 MichNet Joint Technical Staff Meeting
 Michigan State University, East Lansing
 July 25, 1991
 
 Present were: Scott Gerstenberger (Merit/MichNet), Mark Knopper
 (Merit), Larry Blunk (Merit), Ken Josenhans (MSU), Doug Nelson
 (MSU), Tom Hacker (OU), Joel Fletcher (WMU), Bill FIsher (GVSU),
 Mark Strandskov (CMU), George Broomell (WMU), Glenn McGregor
 (Merit/UM), John Vollbrecht (Merit/UM), Allan Rubens (Merit), Gary
 Rosse (UMD), Jeff Ogden (Merit), John Lauro (UMF), Brian Holmes
 (WSU), Paul Schneidenbach (WMU), Eric Sobocinski (Merit/UM), Mark
 Driscoll (Merit/UM), Betty Burke (UMF), Brian Baas (Calvin), Peter
 Murphy (WSU), and Rick McCreedy (WSU).
 
 morning session (recorded by Allan Rubens)
 
 The first order of business was to review the "Merit Dress Code"
 memo, ND 80-999.  A copy of this memo was distributed and reviewed.
 It was decided that the authors of this memo had done well in their
 quest to establish a dress code that would withstand the test of
 time and that very little, if any, revisions were neccessary.  It
 was suggested that the reference to "carriage control" in item 11
 (a situation in which the wearing of sleeping garb is permitted) be
 updated to a more current topic - perhaps "reorganization".  There
 was some confusion as to the meaning of a "mostly bilateral shirt"
 - perhaps this could be shown with a diagram (Nevertheless, all
 attendees did indeed wear mostly bilateral shirts at this meeting -
 Good job!).  The need for an item barring "watches that beep" from
 these meetings is to be taken under advisement by the Dress
 Committee, should such a committee ever be formed.
 
 
 Video Reports (Jeff Ogden, Scott Gerstenberger)
 
 Jeff gave an overview of the last Merit Board of Directors Meeting.
 He indicated that the major purpose of the meeting was to approve
 the budget for the next fiscal year.  The budget discussions went
 quickly.  Increases in member fees of about 5% were approved.  Status
 reports on NSFNET, ANS, and Recruiting were given.  Discussion of
 agreements between the NSF and ANS on commercial traffic were held.
 The Board agreed to allow commercial traffic to flow into MichNet
 from the NSFNET.  A second agreement to allow MichNet to send
 commercial traffic to the NSFNET was reached.
 
 The role of Merit in the area of video was discussed.  Many board
 members feel we should be very aggressive in this area.  There is
 still confusion as to what the board members and others mean by
 "video" and also as to what Merit's role should be.  First the terms
 need to be defined and clarified.  A business plan needs to be
 formed.  Originally $40k was proposed by Merit for a 1 year study to
 come up with this plan.  The Board wanted the plan faster.  They'd
 like the plan and recommendation by early 92 so over $70k has been
 allocated in the current budget to help move this along rapidly.
 Outside consulting will probably be used to come up with the plan.
 Central staff and appropriate Merit member staff will also be
 involved in the video effort.
 
 The first step to complete the business plan by the required date
 will be to find a consultant or consulting firm.  Also people from
 the various Merit members who can contribute to this effort need to
 be identified.
 
 Next Scott discussed Merit's involvement in responding to an RFQ
 issued by the Ingham County Intermediate School District, MSU, and
 Lansing Community College for a video/data network to facilitate
 distance learning.  This network involves the provision of video and
 data services to 44 schools in the Lansing area.  Each school will
 be able to receive 4 incoming video channels and send on 1 outgoing
 video channel.  Thus a given video classroom could allow a class to
 involve 4 remote classrooms.  In addition to the video component of
 this network, a T1 data network is to be provided to all 44 schools
 as well.
 
 Merit was asked by Michigan Bell Communications to assist in the
 specification of the data networking component of their response to
 the RFQ.  We basically proposed a network of cisco routers connected
 by point-to-point T1 channels carried over the same fiber that will
 carry the video channels.  The T1 channels would be allocated on RF
 frequencies between the video channels on the fiber and would
 utilize conventional RF modems.
 
 The proposal was submitted and a presentation was made (after the
 MJTS meeting) and now we and MBC are awaiting notification as to
 whether the bid was accepted or not.  If the bid is accepted, Merit
 may be involved in the installation of the data network components.
 
 Scott also reported on a recommendation formulated by the Merit
 Video Working Group which he chairs that recommends a particular T1
 video codec for purchase by Merit members that will be purchasing
 such devices.  The recommendation is to use the GPT T1 codec.  The
 main reason this particular codec was selected is because it is the
 1st to abide by the CCITT H.261 compression standard.  It is also
 the same codec recommended by MiCTA.
 
 The GPT codec produces "good" color pictures at speeds as low as
 384kbps.  It uses multi-level compression techniques.  Only changes
 from one frame to the next are transmitted.
 
 There is no standard for carrying sound in H.261 as yet.  GPT uses
 16kbps for audio regardless of the bandwidth available.  Thus, about
 128Kbps is the minimal transmission rate that can be used to send
 minimally acceptable video.
 
 Various attendees were asked about their knowledge of video
 activities at their sites.  Joel Fletcher talked a bit about the
 MiCTA trials that have taken place at WMU.  He indicated there was a
 little trouble getting things to work.  Setting up the equipment for
 two way video between different sites is more complex than
 anticipated.  Jeff explained that MiCTA is the Michigan Collegiate
 Technical Association.  It's a group that initially was formed by
 colleges running their own phone systems.  Now they've grown to a
 lobbying and information sharing organization that includes many
 goverment organizations and K-12 schools.  They represent lots of
 buying power to vendors.  Brian Holmes indicated that WSU's
 Engineering school is trying to implement video capabilities for
 their corporate education program.  Scott commented that broadband
 analog video communications only makes sense in cases where a few
 sites need to communicate.  Broadcast channels are too expensive.
 Also it's distance limited because there's too much loss when cable
 systems are strung together to achieve greater distances.  Need to
 go to long-haul analog, satellite, or digital transmission instead.
 
 George Broomell stated that WMU has a group that is heavy into
 video that is totally separate from any data networking group.  He
 wondered whether we should have independent meetings to discuss
 video.  Jeff cautioned against keeping the two groups apart because
 the technologies are coming together these days.  Scott indicated
 that Merit does try to keep in touch with MiCTA to keep tabs on
 what's going on in video in the state.  Someone observed that
 there's already tension between data and telephone groups at many
 schools - video could be yet another group to contend with.
 
 Scott talked about an idea to use the fractional T1 capabilities of
 the new (and some existing) CSU/DSUs that will be used in the new
 Merit backbone.  With this capability 2 channels could be made
 available at each CSU/DSU with no additional equipment - 1/4 T1 for
 video and 3/4 T1 for data.
 
 The Merit central staff may begin doing some experiments with video
 over packet networks.
 
 Tom Hacker said that we should try to get people at various sites
 working on video to work together.  He suggested that a demo of the
 kinds of things that could be done with video these days would be
 good.  Merit should organize such a demo.
 
 John Vollbrecht suggested that Merit might also want to look at
 sending digital signals over the analog channels provided for video
 in cases where sites have video linkages.
 
 
 IP Backbone Statistics Collection (John Vollbrecht)
 
 John described the existing statistics collection capabilities of
 Merit and future plans in this area.  The main goal of this
 presentation was to let people know what we are doing and to
 solicit interest in participating in a working group in this area
 and also to find other applications for the statistics gathering
 tools.
 
 The presentation began by making the point that LOTS of data is
 being collected right now.  Currently work is going on in the area
 of allowing easy access to the collected data based upon a client/
 server model.  A goal is to provide a common set of tools for access
 to the collected data.
 
 The data now collected is from multiple networks: NSFNET (T1 & T3),
 MichNet PCPs/SCPs, Michnet routers, and UMNet routers.  Many kinds
 of data are collected: SNMP, Delay Matrix, NNStat, Off_Net, (e.g.,
 trouble ticket), and configuration.  There are separate collectors
 for each type of data.  The collectors pass the collected data
 either directly to the history data base, or to the data engine if
 the data is in response to a query generated by the data engine.
 The data engine is used by applications to gather current
 information by queries (e.g., SNMP requests).  In addition there's a
 trouble ticket and configuration data base that receives data from
 human collectors.  Applications access the history data base either
 directly (for already archived data) or by the data server when
 using polling to retrieve data when it is presented by the
 collectors to the history db.  Some types of applications allowed by
 this structure are for management reporting, modelling and
 simulation, status and performance monitoring, problem management,
 configuration management, and debugging tools.
 
 For the NSFNET, a delay matrix is formed by pinging from every
 interface to every other interface.  Should this be done for
 MichNet?  (silence).  TCP throughput is measured for NSFNET as well.
 NNStat measures the offered load at all interfaces to the network.
 The offered load is broken dowm into bins according to packet type.
 The Off_Net collector determines what nets are not operational on
 the NSFNET.
 
 There are two machines that run the data collection processes.  The
 stored data is coupled so that the same data is available in both
 machines.  These machines are currently both RS6000s, a model 520
 and a 320.  Hope to be getting two 930s soon which will provide 12Gb
 of disk storage.  All data is archived daily on 8mm tape.
 
 Current applications are: DSU Alert Monitor (only on T3 DSU's now);
 Rover Log Analysis; Monthly traffic summaries, Daily flash reports
 (Trouble tickets).  Other possibilities being considered are for
 link error analysis, Uptime reports, and for traffic analysis.  John
 is quite interested in hearing from others about possible
 applications they might find useful.  He also said that the
 collection facilities could be used to collect data from a given
 sites own routers.  Tom Hacker said that he has a program that looks
 at SNMP stats which might be made to work with this collection
 system.
 
 If you are interested in this area contact John (jrv@merit.edu) as
 he is interested in having a working group meeting soon.
 
 George Broomell is interested in seeing a list of the reports
 available now.  Jeff would like to see a sampler of all the reports
 from which he could say "I want these but I don't care for these."
 Mark Knopper said that his group is looking at unix graphics
 packages to generate nice graphical as well as text reports.
 
 
 Planned circuit upgrades (Scott Gerstenberger)
 
 MichNet has obtained favorable pricing from Teledial.  The existing
 KZ-AA and OK-DT 56kbps links are to be converted to Teledial T1
 links the 1st week of September.  The EL-CMU link will be upgraded
 from 9.6 to 56kbps to provide better IP access to Alma College,
 which is getting an IGS Router.  Before going too far with Teledial,
 we'll wait and see if the circuits really work before doing
 additional upgrades/cutovers.
 
 The EL-GR link is likely to be upgraded from 9.6 to 56kbps to
 eliminate some severe bottlenecking that occurs on the current
 link.  Merit may take over the charges for the AA-DT link and change
 it from MBT to Teledial.  We'll wait until October to decide.  EL-AA
 may be changed to Teledial.  A 2nd 56kbps link may be run to MTU.
 The reason for 2 56kbps links instead of 1 128kbps link is that MBT
 doesn't offer fractional T1 between their POP and MTU (75 miles).
 
 The question was raised about the status of MichNet's outstanding
 proposal to NSF.  Scott said that he'd expect to hear by September
 or October.  He doesn't think anyone has asked the NSF about it for
 about a month.
 
 
 CSU/DSU backbone vendor selection (John Vollbrecht)
 
 The final selection was made among CSU/DSU's from Cylink, DTC, and
 Digital Link.  Digital Link was chosen because of what were judged
 to be superior network management capabilities.  In particular, a
 console on 1 Digital Link CSU/DSU can be used to obtain network
 management info from all the other CSU/DSUs in the backbone.  The
 Digital Link CSU/DSU uses ESF and does the routing of the
 management packets over the ESF.
 
 Six or so of these CSU/DSUs are on order, expected to arrive in
 September.  Peter Murphy pointed out that these CSU/DSUs may not be
 a standard size and therefore might not fit in the available rack
 space at WSU.
 
 
 X.500 Working Group Report (Mark Knopper, also recorded by Mark)
 
 A working group on Directory Services in MichNet met on July 17 in
 Ann Arbor.  Present were: Tim Howes, Mark Smith, Mark Knopper, Brian
 Holmes, Richard Huong, Lih-Er Wey, Rick Schmalgemeier, Tom Hacker
 and Chris Weider.  They discussed the following issues:
 
 --White Pages.  Who is registered now, and what are current
 activities.  How can we get general participation among Merit
 members and affiliates?  What are the problems encountered when
 impelementing an electronic directory, and moving it to X.500?  What
 tools are available for user agents, migration of data, updating.
 How usable is the current software?  What are the existing problems,
 eg. reliability, reliability and reliability.
 
 --Merit Internet directory efforts: the FOX project.  FOX
 participants and their activities: SRI, ISI, PSI, Merit.  Internet
 network infrastructure information in the Directory.  Online
 resources, eg.  library directory.  K-12 people and resources.  NIC
 Profiles.  NSAPs (network service access point addresses), or a
 domain name system for OSI CLNP.  Interoperability between X.500
 impelmentations.
 
 --Statewide activities we might do.  Implement a separate directory
 of MichNet resources or services.  Which Host based user agent.
 Directory of Merit Technical and User Staff people, Merit Board of
 Directories, etc.
 
 Currently UM, MSU, OU and Merit are registered in the PSI White
 Pages Pilot project.  UM has a full database of faculty and staff,
 downloaded periodically from tapes provided by ITD UIS/DSC.  MSU has
 the college of engineering, and will later have the full faculty
 and staff.  OU is currently in transition between machines.  Merit
 has its staff, and a number of other activities.  Specific comments
 on member organizations follow.
 
 MSU - Has separate databases and applications for academic vs.
 administrative staff and faculty.  How to update, and how to
 distribute access are major challenges.  The data on the vax 8700
 currently includes e-mail addresses.  Besides updating, politics is
 also a problem.
 
 WSU - Not currently doing any X.500.  A "finger" service is being
 worked on for the administrative system, which includes user ids
 for the mainframe, which means that it includes e-mail addresses
 for some users.  CIT Planning and support services is getting the
 data from the administrative departments.
 
 UM - The data is now available and being updated.  There are no
 e-mail addresses provided.  Updates occur each month.  Updating is
 tied to the personnel/W-2 database.  There are 25000 faculty/staff
 entries, including 10000 for the Hospital.  This is running on a Sun
 4/470 with 64MB memory and 256MB swap with quipu indexing enabled.
 The process is 31MB memory resident.  The plan is to replicate
 everything on another machine for reliability.  Following that
 multiple machines will be used to allow slave servers for
 sub-organization levels.  DSC claims that 25% of the entries change
 per month! Passwords will be handed out by the ITD business office.
 
 OU - Tom had a DSA running for a while but it is moving to another
 system.  An address registration form used on campus has a question
 about permission to release the information electronically.  The
 2000 faculty/staff will be entered before the 13000 students.
 
 Send questions or comments to mich500@merit.edu.
 
 
 Lunch
 
 afternoon session (recorded by Mark Knopper) WMU Reorganization
 
 WMU has reorganized to combine academic and administrative
 computing.  Joel Fletcher attended the meeting in this newly
 reorganized spirit.
 
 
 SCP Replacement Project (Al Rubens)
 
 Al reported that Merit is talking to Telebit about cooperating in
 development for their terminal server.  They are interested in a
 cooperative effort to make up voids in their existing product.  Al
 is working on a requirements draft.  He needs ideas on needs,
 functions and services for the new SCP.  Other vendors are possible.
 Wellfleet was interested.  Similar devices include Synoptics (note
 they are cooperating with cisco).  An e-mail draft requirements doc
 was sent out by Mark.  It included a writeup produced by Al which
 was slightly modified by Mark.  Do we even need SCP replacements, or
 are vendor offerrings sufficient?
 
 We have 300 SCPs, primarily supporting asynchronous terminal
 support.  Is that needed with move toward workstations?  Answer:
 though advanced workstations get attention there is still a
 requirement for low cost connections.  Scott pointed out there are
 475 dialin ports plus 12000 hardwared lines.
 
 Should other functions be integrated?  Asynch.  Host connections
 along with Dialup are needed (consensus of group).  There is
 currently no growth in number of hardwired ports being installed,
 but dialup and AH are increasing.  SCPs are maintainable, but Unibus
 based PCPs are not due to older hardware.  PCPs must be replaced.
 SCPs are more reliable but won't last forever.  One suggestion has
 been to donate the old SCPs to the K-12 schools as they are taken
 out of service by universities and colleges.  That may or may not be
 practical.  Can we drop some requirements so that users can still
 use SCPs?  K-12 can only use this if it is CHEAP.
 
 We then went through each service and discussed whether it is still
 needed in a replacement product.
 
 Dumb terminal support (VT100: YES.  MCP: NO.  Need ALA for library.
 MSU is using PCtie.  NOTIS is everywhere.  NIMs are harder to
 support.  ISDN Support: Not high requirement but will be in future.
 MBT has a long list of COs that will support ISDN.  Hermes has
 mainly MTS support - MCP and line mode.  We need minimal line mode
 support.  With PPP, we can push line mode editing back into the
 user's workstation.  PPP/SLIP: YES, with TCP Header Compression.
 Hardware flow control is needed to allow this to run over MNP to
 provide modem compression and error correction.  X.25: Requirements
 are dropping in the US.  MSU will turn off most of their X.25
 use--they are retiring the Contel net.  Jeff still gets requests for
 X.25, eg.  large RFPs from state.  X.25 Passthrough: Not needed any
 more.  X.25 Conclusion: NO.  Do X.25 in existing SCPs, don't do X.25
 in routers.  Ferris state wants to use X.25.  Protocol Conversion:
 Telnet to Which host: REQUIRED.  New applications can use tn3270
 rather than conversion in the network.  SCP Replacement MUST include
 tn3270 conversion.  Xyplex supports this.  Routing, ie.  cisco
 trouter: Look at costs and options, configurations.  Two boxes is an
 alternative to this.  Purely dependent on cost.  Host locator and
 Authorization Server: Merit proposes to write a new standard for
 authentication between a terminal server user and an authorization
 server, to be implemented in the SCP and a terminal server.  These
 services are required.  Statistics: YES.  Help facility: Help can be
 provided by a remote system.  No need for special support in a
 terminal server.  Menu facility: Important enough that it should be
 pursued if possible.  Not absolutely essential.  George pointed out
 that the lack of this support in the SCP is why we lost the WMU
 Library contract.  They are using a Dataswitch.  Network Management:
 YES.  Channel Interface: NO.
 
 These are important decisions and will be included in an updated
 SCP Requirements document.
 
 
 Router Administrative Access Policy Issues (John Vollbrecht)
 
 John explained that an exterior gateway protocol between routing
 domains for the MichNet backbone and those of the member campuses
 will be used.  This will be done in part to allow separate
 administrative control for the routers in each domain.
 
 With ciscos, there is only one password allowed per box, with write
 access needed to allow read access.  This is somewhat disappointing
 because we will have to grant write access to a wider group than is
 desired.  In some cases shared ownership is necessary but in general
 the Merit staff will need to have primary access.
 
 John will convene a working group to continue the discussion of
 this.  [List of volunteers provided here.]
 
 
 MichNet Dialin Cost Estimate
 
 Scott and Jeff distributed a spreadsheet on dial in cost estimates.
 They requested that member sites fill in their data and return to
 them.  Merit is looking into alternatives for dialin support, eg.
 replace multiple sequences with 800 access.
 
 
 Working Group Status
 
 The following working groups are in existence. 
 
 Glenn McGregor - authorization
 Leonard Peirce - Usenet
 Scott Gerstenberger - Video
 Mark Knopper - Network Applications
 Scott and Jeff - Dialin
 John Vollbrecht - Router Access
 Mark Knopper - X.500
 
 
 Next Meeting
 
 The next meeting will be on Al's 42nd birthday (October 21), in
 East Lansing.
 
 
 
 
- - - - - - - - - - - - - - - - -