Subsystem specific benchmarks.

Ok, I've uploaded all of the subsystem specific polygraph workloads as
described in a previous mail here (they are documented at the site as
well for those just being brought into the conversation):

http://www.swelltech.com/pengies/joe/squidngbench

I've also provided the scripts I'm using to graph vmstat logs for
inclusion in the polyreport. I haven't automated that inclusion yet,
but might eventually (so for now, if you want them in your reports
you'll have to hand copy the bottom part of one of my index.html files).

Also there is the complete (except for poly3 which is running now and
will be posted tomorrow) set of those benchmarks as they ran on the
reiser_raw branch. I've now updated the reports to include memory usage
graphs, per Sizif's request.

I have my slate full for things to benchmark next, and they will be done
in this order: hno-devel (Henrik's new async code), Nikita's async-tcp
branch, sizif's latest reiser_raw version. Those will get done one per
day, if possible.

I welcome ideas for improvements to the workloads, as they are still
quite simple (mainly a Datacomm-1 load with various modifications).
Polymix-3 proved too hard to accurately modify to suit specific purposes
(I couldn't figure out how to consistently generate the hit ratios and
cachable ratios we needed! ;-).
                                  --
                     Joe Cooper <joe@swelltech.com>
                 Affordable Web Caching Proxy Appliances
                        http://www.swelltech.com
Received on Wed Oct 04 2000 - 01:27:06 MDT