Re: [squid-users] Antwort: Re: [squid-users] memory usage for squid-3.0.STABLE15

2009-07-01 Thread Matus UHLAR - fantomas
On 30.06.09 13:13, martin.pichlma...@continental-corporation.com wrote:
 I checked -- cached objects are not re-checked, at least not with two or
 three hours.
 
 But the memory usage is higher still without icap while the cache is still
 filling -- but this may due to the fact that I configured squid to cache
 objects only up to 1 MB and icap scans larger objects, too.
 
 Additionally squid does not know the icap scan limit, therefore every file
 will be sent to the ICAP server. So the higher memory usage will be
 probably the need of caching large objects, too, at least until ICAP has
 them scanned. Your thought with the two memory buffers may be another
 reason.
 
 OK, thank you, now I understand more clearly the extensive memory usage
 with ICAP. I will have to rethink whether ICAP is really the best way for
 what I want.

Well, I cache bigger files than 1MB, my maximum_object_size is 32 MB now.
with LFUDA policy I found this to have better byte hit rate...

-- 
Matus UHLAR - fantomas, uh...@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
A day without sunshine is like, night.


[squid-users] Antwort: Re: [squid-users] memory usage for squid-3.0.STABLE15

2009-06-30 Thread Martin . Pichlmaier
I checked -- cached objects are not re-checked, at least not with two or 
three hours.

But the memory usage is higher still without icap while the cache is still 
filling -- but this
may due to the fact that I configured squid to cache objects only up to 1 
MB and icap scans larger objects, too.

Additionally squid does not know the icap scan limit, therefore every file 
will be sent to the ICAP server.
So the higher memory usage will be probably the need of caching large 
objects, too, at least
until ICAP has them scanned.
Your thought with the two memory buffers may be another reason.

OK, thank you, now I understand more clearly the extensive memory usage 
with ICAP.
I will have to rethink whether ICAP is really the best way for what I 
want.

Thank you again for your insight!

Martin



On 25.06.09 15:39, martin.pichlma...@continental-corporation.com wrote:
 I have a question regarding memory usage for squid. I have 4 proxies, 
each 
 has about 200-400 req/s and 2-5 MB/s with ntlm_auth and about 1000 lines 

 of acl,
 squid version is 3.0.STABLE15 on Redhat AS 5 Linux.
 They are busy servers and therefore have no disk cache but memory cache 
of 
 6144 MB (6 GB) and provide the internet access for some 10k users.
[...]
 The squid process needed about 7.5 to 7.8 GB and that seems reasonable.
 After we enabled ICAP (c_icap with clamav virus scanning) the memory 
usage 
 of the squid process rose to about 12.8 GB.
 
 Is this a normal behaviour with squid when icap is enabled?

It's quite possible that when icap is enables, squid must reserve some
memory for icap i/o buffers.
Although only one buffer may be teoretically needed, it's possible that
squid uses two of them.

the user memory highly depends on number and size of the uncached objects
being accessed (i think cached objects aren't re-checked, are they?)


-- 
Matus UHLAR - fantomas, uh...@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
WinError #98652: Operation completed successfully.