Poly 2.2.8

From: Alex Rousskov (rousskov@ircache.net)
Date: Mon Dec 27 1999 - 01:11:35 MST


Hi there,

    Web Polygraph 2.2.8 is available. The change log is attached.
        http://polygraph.ircache.net/sources/

    AFAIK, this release does not include bug fixes that are worth
mentioning in the preamble. Most of the changes deal with off-line log
analysis. Lx has been split into lx and ltrace (see change log for
details); both can now handle multiple logs appropriately.

A new tool to generate reports based on binary logs was added
(tools/ReportGen). ReportGen is set of perl- and gnuplot-based scripts. The
scripts generate reports in HTML, including "executive summary" and
"traces" sections. We hope to fully document this tool within 24 hours.

Alex.

---------- change log -------------
2.2.7 -> 2.2.8
19991227
    - added "req_inter_arrival" field to Robot to specify other
      than Poisson inter-arrival time distributions; for example,
      downtime test will use constant distribution (PGL)
    - redesigned lx to concentrate on phase-related statistics and
      support multiple log files; phases with equal names are
      merged, and phases with different names are concatenated to
      produce correct "totals"
    - added `ltrace', a tool to generate traces based on binary
      logs; this functionality was moved from `lx' that no longer
      supports trace extraction
    - changed naming scheme for log objects to be more consistent
      and scalable, added a few new objects
    - log statistics about transactions waiting for resources to
      become available (e.g., requests waiting for a robot
      connection pool to allocate a connection); this changed log
      format, unfortunately
    - dump more information on "foreign content <tag>" and
      "malformed content <tag>" errors
    - added a cool "summary" screen to polymon
    - do not quit when all phases met their goals when there is a
      positive idle timeout; wait for that timeout to expire (so
      that servers do not quit while clients are running)
    - do not log ascii representation of addresses when storing
      current state
    - "level" statistics was not maintained properly between
      phases
    - "fill" stream was measured as "miss" stream instead of
      "cachable miss"; polyfill-2.pg was not affected because all
      polyfill objects were cachable
    - gettimeofday(2) system call is broken on FreeBSD; make sure
      internal Polygraph clocks never go backwards and show sane
      values
    - avoid infinite recursion coredumps when we fail to write a
      log entry; report and try to salvage logging errors instead
    - do not dump core if a phase has no name, use "" name instead



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