Child pages
  • Meeting Minutes from 12-July-2010 in Columbus
Skip to end of metadata
Go to start of metadata

Performance Working Group at Joint Techs in Columbus, OH

July 12, 2010

Attending (partial list):

   Jeff Boote
   Chris Hawkinson
   Carla Hunt
   John Hicks
   Greg Wickam
   Matt Mathis
   Tiziana Refice
   Jose Augusto Monteiro
   Daniela Bauer
   Eric Pouyoul
   Andy Lake
   Aaron Brown
   Brian Tiernery
   Jason Lee
   Yul Pyun
   Seth Perry
   Jon Dugan
   Kavitha Kumar
   Tom Throckmorton
   Dan Pritts
   Heorge Uhl
   Inder Monga
   Katsuhiro Sebayashi
   Jason Zurawski


 Internet2 Update   (Jeff Boote)

- perfSONAR Software
  pSPT 3.1.3 was released in April.  3.2RC1 released today.  CentOS 5.5 based.
  Uses RPMs (all pSPS softare available as RPMs).  Has the web100 kernel
  support on a stock CentOS kernel.  RC, so we need testers.  Also the LiveCD
  and install to disk option.

- New Employee(s)
  Welcome to Kavitha Kumar.  Looking for more help, try to distribute this to
  as many people as you can:

  3.6.3 released.

- Internet2 ION Transition
  Complete.  R&D staff helped move essential services such as lookup and
  topology.  Web interface getting stats via perfSONAR for the junipers

- Standards
  NMC - Targeting a final draft of base doc by Oct 2010
  NM - Editing track for 'base' doc, deadline is Oct 2010
  NML - published a topology doc.  Working on schema doc.

- JET Demo
  Interoperability between different networks.  Bandwidth and Latency testing.
  NASA/NLR/Inernet2/ESnet/UEN/NOAA.  Come to JET meeting for more.

Network Diagnostic Tool and the API from the Virginia Tech Perspective

Seth Peery, Virginia Tech Geospatial Information Sciences


   Basic idea of talk - present the NDT JavaScript API.  Talk about why it was
   needed and what it is good for.  Also where the discussion on NDT needs to

   Wanted to originially develop a framework for linking NDT with other web
   apps.  Our use case was for mapping - e.g. Seth is in charge of GIS data at
   the university.  Wanted to be able to map things like the performance
   measurements and geo-locate the sources.  Useful for making things like a
   community broadband map - shows the VT example (google map of test results).

   How did we get there?  Initially added some simple hooks to NDT (its open
   source), set up a server and had people test to it.  Couldn't start with a
   whole lot else (e.g. tools like ookla are closed source).  Started slow,
   simple upload/download hooks.  The changes were made to the applet portion,
   then applications can be developed against deployed servers. Others thought
   this was a good idea, then came to ask for more variables, and more formal
   treatment, e.g. MCNC, Broadband Census, Virginia Government.

   Measurement lab approached as well - were doing work with the FCC.  With the
   changes to the applet being targeted as a way to make new applications (that
   hid details of NDT).  Hardened the changes to the API, made it a little
   nicer and documented it.  Other MLab bits - deployed infrastructure, using
   tools like DONAR to help users test to a close server.

   Getting other apps to use the API - simple code import (see web above).
   Creates new 'front ends' and at the same time each test will record tests
   for MLab. 

Q: All the variables from web100 exposed? 

A: Not all - popular ones that made sense.  Can always add more.

   Implications on 'what have we done' - NDT is a diagnostic tool that tells
   lots of things.  Does adding it into a simple 'speed test' make sense?  Does
   it do this well enough?  Read recent MIT report and complaints when compared
   against something like Ookla.  Basically bringing a research tool for one
   purpose into the 'mass market' speed test arena.  Is this the right

Audience feedback:

   Chris H: NDT is the tool we turn too first for debugging things and getting
   a good baseline.  Easy to run (for users), rich results to go though.  After
   getting started, migrate to new tools for the most part.  NDT is a 'tcp
   performance tester', wouldn't think it does well at raw speed tests.

   Brian T: Research community will benefit from the data that mlab is
   collecting, even if it is 'flawed' as pointed out by MIT.  Ways to show that
   the data is still useful, error estimations, etc.  Also some thoughts on
   if TCP reno (used in NDT) is good for measurements compared to other

   Some talk about what MLab is doing - e.g. Big Query (see talk later this
   week on it).  Seth brings us back to the question - is NDT the right tool

   Jeff B: This WG should be used to gather requirements.  What should happen
   to NDT?  It needs a lot of maintenance to keep it relevant.  E.g. what
   happens in 100G testing?  Heuristics need work, need to answer the
   congestion control question.

   Matt M: Brings up some old IETF knowledge.  E.g. IPPM wanted a simple way
   to test capacity.  Knew it was hard.  In NDT's case it has a repeatability
   problem.  Works well at what it does (being able to tell very well what TCP
   is doing).

   Seth: Wants to see someone from R&E (someone who will be more open) be the
   standards bearer for consumer broadband testing.  Do need to make sure that
   NDT doesn't fall out of favor and not be able to go forward in the space.
   Notes that tools like ookla work well at what they do, we just don't know
   how they work...Would rather an imperfect tool that admits its flaws and
   people can see them vs an imperfect tool that doesn't admit failure and
   doesn't allow people to see what is going on.

Presta 10G Update   

Come see demos at SC10

Discussion on pS/Performance Directions - (Jeff Boote)

   We spent a long time making perfSONAR stable.  Where do we go from here? What else
   does the community want?  Perhaps: netflow? auth? closest MP

   Do we need a task force from this group ?  Will be summarizing the
   results of the July 8-9, 2010 perfSONAR workshop soon.  That could be a good indicator of what is

   Some discussion on the NOC process - and how integrating the tools is still
   a work in progress, but happening.  Not enough representatives from an
   end of line network (campus) in the room.  Some thoughts from an exchange
   point (latency is the enemy, little problems made worse).

   GridFTP work - expose the results of tests.  Limit active testing?
   Web100 update later in the week, to see where that is going.  Still valuable
   for what it can expose.  Tools use it.

   A question on 'layer 2 visibility'.  Also some clarification on what pS is
   (a data model and schema) and what is isn't (re-making a lot of performance

  • No labels