Re: FW: Request for CVS Branches.

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Fri, 10 May 2002 22:41:30 +0100

Great!

You have hereby been given developer access to devel.squid-cache.org.
Please take some time to read the documentation on how to proceed (if
not, you mainly risk loosing your own edits..)

Required tools and howtos: http://devel.squid-cache.org/CVS.html
Some basic rules: http://devel.squid-cache.org/rules.html

If you have any questions, please ask.

Regards
Henrik

On Friday 10 May 2002 11:08, squid-dev@visolve.com wrote:
> Hello Henrick / Adrian
>
> We failed to mention in our previous mail given below that we also
> have a sourceforge account with account name "visolve"
>
> Expecting to hear from you on the CVS branches.
>
> Regards
> Squid team.
> www.visolve.com
>
> -----Original Message-----
> From: Squid Developers @ Visolve.com [mailto:squid-dev@visolve.com]
> Sent: Friday, May 10, 2002 1:06 PM
> To: adrian@creative.net.au; hno@marasystems.com
> Cc: squid-dev@squid-cache.org
> Subject: Request for CVS Branches.
>
>
> Hello Henrik / Adrian,
>
> This is regarding some of the development work in Squid undertaken
> by ViSolve team. We have partially completed the following two
> implementations.
>
> 1. RT Signal implementation.
> The objective of using RealTime signals to implement network I/O is
> to reduce CPU usage dramatically, thereby improving performance.
> The select/poll model must scan through a huge array of file
> descriptors to see which ones are ready for reading or writing.
> This scanning takes up lots of CPU time. With RealTime signals,
> read/write events for a socket are placed on a queue. To retrieve
> events, instead of polling, one just retrieves the first item in
> the queue.
>
> We have tested squid-2.4.STABLE3 and squid-2.4.STABLE3 with RT
> signal patch by disabling the disk. The results show very
> sigificant improvement in the performance. These two tests are
> taken with polymix-3 and hardware configurations are
> Processor : Pentium-3 933MHz
> RAM : 512MB
> NIC : 100Mbps card
>
> Our RT signal squid went upto 500 req/sec with CPU usage is below
> 80%, where as squid-2.4.STABLE3 reaches 100% CPU usage in 150
> req/sec itself. When we test our RT signal code by enabling the
> disk It is satisfying upto 230req/sec in the above configuration.
> During the test the CPU usage is less than 40% and memory usage for
> squid is 105MB, overall response time is 1.3 sec, miss response
> time is 2.5 sec and hit response time is .26 sec. This is 55% more
> than the normal squid.
>
> We have implemented this in Linux OS, kernel 2.4.13 by adding a
> patch for RT Signal. We have also identified other challenges which
> we think could be overcome with your help.
>
> 2. WCCP 2.0 compliance.
> Reference: Open source code developed by Joe Cooper in Squid 2.2
> for Wccp 2.0
> compliance. We have worked on the above said reference and have
> changed Squid 2.4 and
> Squid 2.5 for WCCP 2.0 Compliance.
>
> We request for 2 separate CVS branches for these codes or request
> suggestions from you on the
> procedure to submit the code and to share our futher developments
> in these two branches
> with Squid developers who would be interested to further this
> effort.
>
> Best Regards,
> squid team
> www.visolve.com
Received on Fri May 10 2002 - 16:01:49 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:15:25 MST