Delay pools and large files

>> Last year (11.11.98), Martin Bene wrote about "Delay pools setting",
>> I think that his mail wasn't answered, and I'd like to know
>> which squid version will fix this problem?

> I think I answered the e-mail at the time. The problem is the way squid
> scans filedescriptors. It would be possible to make squid behave "better",
> for example alternating through a set of different scanning orders. Only
> delay pools are affected, and the impact is relatively minor in the real
> world, so I haven't looked into it very far. I did suggest some
> simple changes
> which could be made but didn't work on them yet. I think this will probably
> be a 2.3 thing (since most likely nobody else will work on it, and I'm
rather
> busy at the moment) - I'll try and make a patch within the next 3 to 6
months.

From what I've seen impact can be quite severe - enough so that I'm now
just using a traffic shaper to limit bandwidth served to clients instead of
just limiting the cache-misses. While OVERALL performance is wores for the
client(s), INTERACTIVE performance is much better. This is especially true
if I'm not limiting single useres (those I could argue with :) but
firewalls/ proxys for subnets where the whole subnet can have, say, 10
users not getting any data because of one download.

Impact mostly seems to depend on how busy your uplink is - the better your
connectivity, the higher the chance that you run into this "delay pool
saturation" effect.

So Benarson isn't the only one silently praying that you get around to
addressing the issue :) I've tried looking into the sources to see if I can
fix it myself but haven't had much luck so far; if you could point me in
the right direction maybe I'll be able to do this myself.

Thanks again for your work on this great program,

Martin.

"you have moved your mouse, please reboot to make this change take effect"
--------------------------------------------------
 Martin Bene vox: +43-664-3251047
 simon media fax: +43-316-813824-6
 Andreas-Hofer-Platz 9 e-mail: mb@sime.com
 8010 Graz, Austria
--------------------------------------------------
finger mb@mail.sime.com for PGP public key
Received on Tue Apr 27 1999 - 10:20:03 MDT