menavaur/Autobench

Autobench - an HTTP benchmarking suite

Copyright (C) 2001-2003 Julian T. J. Midgley <[email protected]>

Homepage: http://www.xenoclast.org/autobench/

Licence
-------

Autobench is distributed under the terms of the GNU Public Licence, a
copy of which may be found in the file "LICENCE".

Manifest
--------

The 'autobench' package comprises the following programs:
  
   autobench
     - A command-line wrapper for httperf which runs httperf a number
       of times against a target host, producing a CSV or TSV format
       file for later analysis/graphing.

   autobench_admin 
     - The administrative controller for distributed autobench.  Used
       to issue commands to a cluster of machines running autobenchd,
       in order to conduct synchronised multiple-client benchmarks
       against a target host.  Collates the results from all the
       clients in the cluster, and outputs CSV or TSV format file as
       with autobench.

   autobenchd
     - The client dameon for distributed autobench.  Listens for
       commands from autobench_admin, and runs httperf to benchmark
       the target host.  Summarises the results and returns them to
       autobench_admin for collation.

   bench2graph
     - Takes TSV benchmark results files produced by autobench or
       autobench_admin and plots them using gnuplot.

   crfile
     - small utility for rapidly creating files filled with random
       data of a specified size.  Useful for creating data sets for
       benchmarking.

   sesslog
     - Creates session logs for use with the --wsesslog argument of
       httperf from NCSA Combined or NCSA Common Log Format web access
       logs.  Useful for generating real-life workloads from your log
       files.

Man pages are included for all of the above.

Installation
------------

See the file "INSTALL" for installation instructions.

Use
---

Autobench can be used in one of two modes.  In the non-distributed
mode, a single autobench client runs benchmarks against one or two
servers.  See the "autobench(1)" man page for details of this mode.

The distributed mode allows a server to be benchmarked using several
clients simultaneously (reducing the risk that the clients themselves
will become the bottleneck).  See the "autobench_admin(1)" and
"autobenchd(1)" man pages for details.