Poly 2.2.6

From: Alex Rousskov (rousskov@ircache.net)
Date: Tue Dec 07 1999 - 09:53:11 MST

Hi there,

    Web Polygraph 2.2.6 is available. The change log is attached.

    This release includes bug fixes and configuration improvements. We
were hunting the size distribution bug for quite a while and finally got
it fixed. Testers should expect smoother and more predictable workload

    As you know, we have finally committed to do WAN simulation during
the second bake-off. However, no Polygraph changes relevant to that
decision were necessary. Unfortunately, the piper tool cannot be used
"as is" at the moment because we had to reduce the number of pipes to
two per machine (from two per IP alias; due to DummyNet performance
limits). We have posted ipfw configuration commands required to
configure DummyNet pipes. We will try to tune piper so it can still be
used in this "simplified" case.

    We ran out of time and did not add the final bits of "file
extensions" support required for L7 switch tests (requested a few days
ago on this list). Sorry. Will shoot for the next release.

    Our TODO lists includes logging/monitoring improvements and double
checking two pending assertion reports (we may have fixed the bug, but I
am not 100% sure since the bug is not reproduceable).


---------- change log -------------
2.2.5 -> 2.2.6
    - PolyMix-2 workload now uses 0.4/sec per-robot request rate; use
      cloning or IP aliases to get more robots (and hence higher
      request rates)
    - added "rep_sz_mean" object to lx
    - added polyfill-2.pg to workloads/
    - added pmix2 blob to tools/BB/bb.pl and polished bb.pl's comments
    - when using --dump, dump transaction meta-info even if there is
      no message data to dump
    - recognize KB scale for command line size options
    - two meant-to-be-independent random number generators used for
      simulating object sizes got "in-sync", producing wrong size
    - phases with decreasing request rate were dropping request rate
      too fast
    - public_interest for PolyMix-1 and DataComm-1 workloads should be
      set to 0%, not 50%.
    - added now-required hit ratio and popularity model specs to
    - avoid division by zero on corrupted log files

This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:10 MDT