RE: Polygraph 3.0.5 available

On Wed, 2006-09-27 at 15:32 -0400, Charlie Younghusband wrote:

> Have you (or anyone else on the list) folks given any thought to modifying
> the default content simulation models to reflect the 'modern' web sites?
> Such as the fact that more web sites assume users have broadband and thus
> more graphics as well as changing web presentation technologies, compared
> the last time this was touched, probably > 5 years ago...

I agree that "modern" distributions of web object sizes and the number
of embedded objects may differ from what we have observed and agreed on
a few years ago. The skeptic in me doubts there have been drastic
changes though.

Standard workloads used to mimic a "very busy" web client accessing
"busy" sites. These days, same clients may be considered "normal" or
"average", and I do not think busy site response times decreased
dramatically.

PolyMix and WebAxe workloads simulate embedded objects. I would not be
surprised if the average number of embedded objects per "container" have
increased (PolyMix-4 is using zipf(13) distribution for that number,
yielding a 4.5 mean). You are certainly welcome to tweak the workloads
to reflect whatever you think a reasonable number is.

As for "changing web technologies", it is not clear to me whether they
have relevant effects beyond the parameters discussed above. If a
particular HTTP intermediary looks at response content, Polygraph
content simulation module can be used to replicate real content:

  http://www.web-polygraph.org/docs/userman/csm/

Finally, Polygraph 3.0 contains a set of scripts that can derive
workload parameters from an access log file. This lets you tailor a
workload to what _you_ believe is a representative traffic sample.

  http://www.web-polygraph.org/docs/userman/access2poly.html

A lot more can be done in this direction. Contributions and work
requests are welcome.

Thank you,

Alex.


> -----Original Message-----
> From: users-owner-mN9fGWdlm5pku/f+YMZH/di2O/[email protected] [mailto:users-owner-mN9fGWdlm5pku/f+YMZH/di2O/[email protected]]
> On Behalf Of Alex Rousskov
> Sent: September 27, 2006 11:01 AM
> To: Polygraph Users
> Subject: Polygraph 3.0.5 available
> 
> Hello,
> 
>     A new major public release of Web Polygraph is available after more
> than a year-long stretch of private development snapshots.
> 
>     http://www.web-polygraph.org/downloads/
> 
> Major v3.0 change include:
> 
>   - HTTP pipelining support [1]
>   - HTTP compression support [2]
>   - HTTP Cookies support
>   - GCC 4.x support
>   - Apache License without publication restrictions [3]
> 
> Please see the change log[4] for a detailed list of improvements. Most
> of the new features are documented. Upgrading from v2.8 should not cause
> major problems as workload defaults remained the same.
> 
> This release may not be stable yet, but its isolated features have
> received considerable amount of testing from customers with access to
> development snapshots.
> 
> The Measurement Factory resources and development contracts with
> FrontPorch, Network Appliance, WildBlue, and XipLink (among others) for
> the past two years have made this release possible.
> 
> Enjoy, 
> 
> Alex.
> 
> [1] http://www.web-polygraph.org/docs/reference/models/traffic.html
> [2] http://www.web-polygraph.org/docs/userman/compression.html
> [3] http://www.web-polygraph.org/downloads/copyright.html
> [4] http://www.web-polygraph.org/downloads/changelog.html