Polygraph 2.7.2 available

From: Alex Rousskov (rousskov@measurement-factory.com)
Date: Mon Aug 13 2001 - 16:42:55 MDT


Hi there,

    Web Polygraph 2.7.2 is available. The change log is attached.
        http://www.web-polygraph.org/downloads/

    This is a minor release incorporating PolyMix/WebAxe-4
improvements, new Byte Hit Ratio simulation algorithm, and bug fixes.
Please see change log for details on these and other changes.

    All 2.7.0 and 2.7.1 users are encouraged to upgrade for many bugs
have been fixed since those versions were released.

Alex.

------------- change log -----------------
2.7.1 -> 2.7.2
20010813
    - calculate values for If-Modified-Since header fields using
      Object Life Cycle configuration of the corresponding object
      content type instead of using cached last-modified values;
      robots could not cache enough objects to generate reasonable
      number of IMS requests with "sane" IMS headers; the new method
      solves the problem and eliminates requirement for maintaining
      a private cache on each robot (saving lots of RAM for large
      scale configurations); now we can specify the number of
      IMS-200s and IMS-304s and should be able to meet the specs;
      this is a yet another change that makes it crucial to have
      identical client- and server- side PGL configurations
    - select object birthday as a random point within the object
      life cycle starting at "time zero" instead of allowing
      relative or absolute birthdays; birthday point should not
      matter for any current practical purpose, and having relative
      birthdays made it impossible(?) to re-generate OLC timestamps
      in non-authoritative processes (e.g., clients); the latter was
      needed for robots to generate sane IMS requests; PGL object
      birthday settings are now deprecated and ignored
    - removed private_cache_cap settings (i.e., disabled robots'
      private caches) from all standard workloads; we no longer
      need a private cache to generate IMS requests and, hopefully,
      the latter was the only reason to have it; private caches
      consumed a lot of memory on large scale configurations
    - added "bhr_discrimination" field to the PopModel type to
      control the aggressiveness of the byte hit ratio (BHR)
      discrimination algorithm; 100% makes algorithm the most
      aggressive (default); 0% disables the algorithm; internally,
      bhr_discrimination is actually a probability of applying the
      discrimination algorithm when oid selection is made
    - added bhr_discrimination to generation-4 workloads
    - make request body size of PUT and POST requests random rather
      than 8192 bytes; 8192 bytes is now an upper limit given 16KB
      I/O buffers
    - moved WebAxe-4 details into include/webaxe-4-guts.pg and
      synchronized WebAxe-4 specs with recent PolyMix-4 changes to
      keep the two workloads similar (needs more work)
    - removed "recurrence" field from content types in contents.pg
      since Polygraph does not use it anymore; warn if "recurrence"
      is set
    - increased public_interest from 50% to 75% because we now have
      robots that are not active until after the [private] working
      set size is frozen and, thus, some robots will not be able to
      request objects from their private working set, violating the
      old goal of 50% of request submitted to private subsets;
      while no drastic changes are expected due to this change, we
      need more measurements to understand its effect
    - removed birthday settings from standard OLC configs; OLC now
      ignores birthday settings
    - decreased the ratio of IMS requests in PolyMix-4 from 20% to
      15%
    - configured PolyMix-4 robots and servers to abort 0.1% of
      transactions
    - added 1.5% of POST and 0.1% of HEAD request methods to
      PolyMix-4 and WebAxe-4 workloads
    - added zone name back to AddrMap in PolyMix-4; dns_cfg should
      work with polymix-4.pg now
    - increased supported and required log version to 10 due to
      transaction logging changes; older logs cannot be read with newer
      readers and vice-versa
    - IpsToNames() now preserves port numbers (PGL)
    - supply more information about DNS timeout errors
    - DNS answers were allocated but not freed, leaking a memory on
      every DNS request
    - try other DNS servers (if any) when DNS query times out
    - cloned addresses are now iterated depth first when agent
      aliases are created, because alias creation code tries to
      allocate n-th chunk of addresses rather than every k
      addresses; the old code would create twice as many aliases for
      addressing schemes with more than one agent per address; the
      new code creates the same alias twice; needs more work to
      avoid extra creation
    - "--names" option was treated as "--addresses" option (dns_cfg)
    - improved old IP alias deletion on OSes with numbered virtual
      interfaces (e.f., Linux and Solaris); this needs more work;
      for example, one cannot have both polyclt and polysrv creating
      aliases on the same Linux box
    - fixed netmasks and broadcast address settings on systems with
      "numbered" virtual interfaces and without ifaliasreq structure
      (e.g., Linux and Solaris); old code did not set netmasks and
      broadcast addresses at all (effectively), and defaults were
      probably assigned by OS
    - when creating aliases, we were opening a new socket instead of
      closing an old one, creating thousands of "extra" sockets when
      aliases were created on a box; this led to "too many open
      files" errors on some systems, at least



This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:19 MST