On Thu, 24 Aug 2000, you wrote:
> > If that's the case (and if it isn't, I'd like to find out what it is), what can
> > I do to fix it?
>
> For an accelerator you probably want to remove the request windowing,
> and instead have Squid suck down the reply as fast as possible from the
> origin server and then push it out to the clients at their speeds.
>
> change src/forward.c:fwdCheckDeferRead to always return 0.
Yes, that works quite well.
> Note that there is an issue in memory usage if you have very large
> objects.
A question on that. The binary server's memory looks like this:
Memory usage for squid via mallinfo():
Total space in arena: 174075 KB
Ordinary blocks: 51552 KB 6714 blks
Small blocks: 0 KB 0 blks
Holding blocks: 176 KB 1 blks
Free Small blocks: 0 KB
Free Ordinary blocks: 122523 KB
Total in use: 51728 KB 30%
Total free: 122523 KB 70%
Memory accounted for:
Total accounted: 48374 KB
Memory pools are limited to 20MB and, looking at the Memory Utilization, that's
all it has for currently idle. It simply isn't releasing any memory.
This was using whatever malloc library came with Debian Linux 2.1. Is that a
bug (would Doug Lea's library release that 122MB back to the system?) or
is squid intentionally holding it for something?
Thanks
-- Brian
-- To unsubscribe, see http://www.squid-cache.org/mailing-lists.htmlReceived on Fri Aug 25 2000 - 00:02:53 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:55:05 MST