OpStat Minutes

[Henry Clark <[e-mail]>]

  Minutes of the Operational Statistics Working Group
  Danvers, 1530 on Mon 3 Apr 95
  
  Agenda
  ======
  
     Introduction
     1404 Update
     Client-Server Draft
     SNMP User's Guide
     Fractals: Traffic and Simulation
  
  1404 update
  -----------
  
  The 1404 draft is currently out as an I-D.  One final change will
  be made to the BNF to slightly change the LABEL section to make it
  visually the same as other sections by re-arranging fields within
  the LABEL section.  The last section (which briefly considers variables
  in other MIBs) will also refer to the Accounting Meter MIB.  The new
  version, along with some assorted editing corrections, will be re-submited
  as an I-D and should goto the RFC editor as an informational RFC by the
  next IETF.
  
  Client-Server Draft
  -------------------
  
  The client-server draft has incorporated all the changes submitted
  on the mailing list (essentially changes to make the text clearer).
  Two changes were discussed and consensus was reached on incorporating
  them at the meeting:
  
  1)  to change the LOGIN command to provide for stronger authentication
      with a challenge-authentication response, which might look like:
  
      >LOGIN "henryc" "skey"
      <CHAL "73 lo39087"
      >AUTH "COW DOG BARK CAT BULL DAY"
      <OK
  
  2)  change the messages returned by the server from the ERROR/OK to
      more like the numbers returned by SMTP and FTP, as in:
  
      101 Login OK
  
  H. Clark will update the draft and resubmit the I-D.
  
  SNMP User's Guide
  -----------------
  
  Work is proceeding at a disapointingly slow pace.  This document will
  contain general recommendations and experiences about how SNMP and
  the specific MIB variables can be used to manage a network.  Some work
  is starting within the Network Management area, which was described
  by Bill Norton, to add a document with every MIB describing the use of
  each variable and algorithms for using the variables to manage a network.
  Our document wouldn't be a standards track document, unlike the SNMP MIB
  documents.  F. Serr will forward some text to the list which might be the
  start of such a document, and the chairs will compile the document (along
  with text contributed by WG members) into  the final form.
  
  Fractals: Traffic and Simulation
  --------------------------------
  
  N. Brownlee discussed traffic studies that had been done at his site
  and the statistical analysis of said traffic.  They discovered that
  the traffic had a self-similar nature; that is, the graphs looked the
  same irrespective of the period of measurement.  This means that you
  will see traffic peaks of any chosen value, if you only wait long enough.
  
  Studies were also made of the variance of the data when aggregated in
  groups of m measurements;  it was observed that the traffic had variance
  decaying slower than 1/m.
  
  Simulations were performed utilizing distributions with finite mean and
  *infinite* variance, such as Student t2, which seemed to produce plots
  similar to the observed plots.  More work needs to be done to determine the
  characteristics of the traffic in more detail, and further presentations
  will be made as appropriate.