1. 28 Jun, 2006 1 commit
  2. 27 Jun, 2006 1 commit
    • Mike Hibler's avatar
      Enough already Mike! A couple of changes for high latency (> 2 sec) links. · bf403713
      Mike Hibler authored
      Change the rude.cfg script we generate to have a final "wait around" period
      (MODIFY with 0 rate) before exiting.  We have it wait for the one-way latency
      of the link.  This way, rude will not exit--and linktest will not do the
      barrier synch--til all packets have had a fair chance to get to the other side.
      This way we don't need a randomly chosen sleep after the barrier returns
      and before we read the crude results.  Though rude already supported the
      MODIFY with zero rate, I had to fix it since it didn't work if it was the
      last command for the flow.
      
      Adjust the timeout period for ping when doing the latency tests.  Previously
      we waited for 3 seconds, now wait for max(RTT+1, 3) seconds.
      
      Much more work will have to be done to support multi-second latencies
      all around, this was just the low-hanging fruit.
      bf403713
  3. 20 Jun, 2006 3 commits
  4. 19 Jun, 2006 1 commit
  5. 16 Jun, 2006 1 commit
  6. 13 Jun, 2006 3 commits
  7. 12 Jun, 2006 1 commit
  8. 08 Jun, 2006 1 commit
  9. 07 Jun, 2006 2 commits
  10. 31 May, 2006 2 commits
    • Mike Hibler's avatar
      Use the new Utah-added -A option to iperf to attempt to fine tune the · e1cbdaaa
      Mike Hibler authored
      FIN timeout thus getting more accurate BW measurements.  From the code:
      
                        #
                        # Even without packet loss, there is a good chance that
                        # iperf's end-of-stream (FIN) will be lost since we are
                        # over-driving the link.  The ACK for that FIN needs to
                        # be received before iperf's timer stops, so a long
                        # timeout (for resending the FIN) on a short duration
                        # run will lower the BW reading considerably.  So we
                        # attempt to keep the timeout as short as possible,
                        # taking into account the round-trip latency of the link.
                        # We may also need to adjust the run time of the test
                        # upward, but we do not exceed the indicated max duration.
                        #
                        # 250 ms is the default timeout value for iperf.  We
                        # start with a candidate value of 50 ms which is chosen
                        # as it is sufficient for the 3% BW error on min duration
                        # (3 second) tests.  You should not pick a base acktime
                        # less than the resolution of the clock (10ms or 1ms).
                        #
      e1cbdaaa
    • Mike Hibler's avatar
      Add new '-A <timo>' option to set the length of time the client will · 95f14c95
      Mike Hibler authored
      wait for an ACK of its FIN.  The fixed 1/4 second was too big for us.
      The option gets passed over to the server when running in tradeoff or
      dual modes.
      95f14c95
  11. 25 May, 2006 2 commits
  12. 19 May, 2006 1 commit
    • Kirk Webb's avatar
      · e0c33c75
      Kirk Webb authored
      More reliability fixes/enhancements.  Hopefully this will catch the remaining
      problems I've seen happen.
      e0c33c75
  13. 05 May, 2006 1 commit
  14. 03 May, 2006 1 commit
  15. 02 May, 2006 2 commits
    • Kirk Webb's avatar
      · e5e0a604
      Kirk Webb authored
      Fix up last change to makefile to actually do the right thing...
      e5e0a604
    • Kirk Webb's avatar
      · 7db226f7
      Kirk Webb authored
      Commit perl event wrapper code generated by the new Swig 1.3.  Also fix
      up the makefile to do a better job of detecting the perl version when
      compiling these.
      7db226f7
  16. 26 Apr, 2006 2 commits
  17. 25 Apr, 2006 1 commit
  18. 20 Apr, 2006 1 commit
  19. 19 Apr, 2006 1 commit
    • Kirk Webb's avatar
      · a5ea34cb
      Kirk Webb authored
      Update the plab proxy to manage restoring the connection to the remote
      elvind on ops itself (instead of leaving it up to elvin).
      
      This whole program needs cleanup/refactoring attention.  Later.
      a5ea34cb
  20. 18 Apr, 2006 2 commits
    • Kirk Webb's avatar
      · fb37c449
      Kirk Webb authored
      Look at "SCHEDULER" attribute to get most everything.
      fb37c449
    • Kirk Webb's avatar
      · e219aa4d
      Kirk Webb authored
      Added silly little program to dump the set of subscriptions registered
      on an elvind server.  Uses the "quench" feature to grab this set.  Note
      that there is no way to just ask for everything, you have to specify the
      subscription attribute you are interested in.  However, since the testbed
      elvin library wrapper always adds certain fields, just asking for one of
      these almost certainly gets you everything you are looking for.  The only
      exception (missed items) would be custom subscriptions created directly
      via the elvin library that do not include one of these common fields.  I
      don't think such subscriptions are ever created in practice.
      e219aa4d
  21. 12 Apr, 2006 2 commits
  22. 06 Apr, 2006 1 commit
    • Mike Hibler's avatar
      Subtle change to the query that checks whether a node's OS supports linktest: · e19449fd
      Mike Hibler authored
      change the query to check nodes.node_id instead of nodes.phys_nodeid against
      the reserved.node_id.
      
      This means the query now returns FBSD-JAIL for vnodes rather than NULL
      for the OS.  I ran the query against every current experiment in the DB
      and the only other changes were:
      
        * plab nodes return RHL_PLAB instead of NULL.  As long as linktest is
          not an osfeature of that image (it isn't) then this fix is compatible
      
        * pc2000s, which host IXPs, now only return one row in the query.
          Previously they would return a row for "pc17X" and for "ixpN".
          This is probably not what we wanted anyway.
      e19449fd
  23. 28 Mar, 2006 1 commit
    • Kirk Webb's avatar
      · 30ff680b
      Kirk Webb authored
      Let's increase the idle check period a bit...
      30ff680b
  24. 27 Mar, 2006 1 commit
    • Kirk Webb's avatar
      · 74fc532b
      Kirk Webb authored
      Some event system fixes for the plab proxy and testing enhancements to tbsend.
      
      * event.c / event.h
      
      Fix assertion failure that appears to be caused by notifications coming in
      on a subscription that is pending deletion (via async API call).
      
      Also add idle polling to the ops endpoint connection to help maintain
      connectivity (will do failover).
      
      Add function to set failover flag on connection handle.
      
      * tbsend.c
      
      Add ability to set connection retry, failover, idle polling, and perform
      looped and triggered testing.
      74fc532b
  25. 13 Mar, 2006 1 commit
  26. 09 Mar, 2006 1 commit
    • Sachin Goyal's avatar
      · 2ca5a9e4
      Sachin Goyal authored
      Modifications done for plab scheduler:
      * pass thru the UPLOAD/CLEAR events to local elvind so that
        plab-scheduler can get them.
      * filter events with <plabsched, 1>
      * listen to scheduler==2 evennts also from ops elvind
      2ca5a9e4
  27. 03 Mar, 2006 1 commit
    • Timothy Stack's avatar
      · 66ee32fc
      Timothy Stack authored
      Clear out plab evproxy subscriptions when the event scheduler is stopped.
      
      	* event/sched/event-sched.c: Add an __ns_teardown sequence that
      	can be used to send events when the scheduler is stopped.
      
      	* event/tbgen/tevc.c: Add a timeout flag that can be used to bound
      	the time spent waiting for an event to complete.
      
      	* tbsetup/eventsys.proxy.in: When stopping/replaying, run the
      	__ns_teardown sequence and wait for it to complete.
      66ee32fc
  28. 02 Mar, 2006 2 commits