wojtek@3miasto.net wrote:
>>>partition for OS, logs, small swap etc and 28GB for squid.
>>>
>>
>>This will handle about 110 reqs/sec, and about 10Mbps total throughput
>>(or about 7Mbps actual uplink bandwidth) in a Linux and Squid+async i/o
>>configuration. On FreeBSD with DiskD this will yield about 80-90 reqs/sec.
>>
>
> PER DRIVE!!!!
Sorry, but no. You might see 70-80 from the first IDE 7200 RPM drive.
30 from the second. 15 from the third...and so on. 10k RPM SCSI drives
raise this by a small amount, 15k drives by another small amount. Squid
does not scale linearly with number of drives. Many, many benchmarks
will bear this assertion out...there is /no/ point arguing about it.
Obviously the other hardware in the system can help squeeze more from a
number of drives (more RAM is very helpful--which is why I always
configure RAM very generously in our systems).
Duane's Squid+FreeBSD box did 140 reqs/sec (am I recalling correctly,
Duane?) at the most recent cacheoff with 6 SCSI 10k drives and 768MB of
RAM. We took a 3 x 7200 RPM IDE box with 512MB of RAM, and did 120
reqs/sec on a Linux+ReiserFS box. Note that three drives did /not/ do
110 reqs/sec per drive or 330 reqs/sec total, or anywhere near it. And
Duane's BSD box did /not/ do 660 reqs/sec from his 6 SCSI disks.
>>Silly anti-Linux FUD. Linux and Squid+async i/o on ReiserFS is the
>>fastest Squid configuration you'll find, and our boxes configured this
>>
>
> i've used linux many year, reiserfs about 1 year and yes reiserfs is
> fast, but too unstable and buggy.
As I've said, I have caches that have been running for over a year on
ReiserFS. Not one FS crash yet on about 30 heavily loaded machines at
ISPs plus several local test boxes...Chris Mason killed the last crasher
bug for Squid loads in the kernel 2.2.16 version of ReiserFS over a year
ago, just before the second cacheoff. So, sorry, again this is entirely
unfounded FUD.
ReiserFS has had some major bugs dealing with NFS, and some other weird
issues...but Squid running on ReiserFS has been very stable for quite
some time (I'm the one who found most of the Squid triggered bugs--I
know what they were and I know they are gone now). The vast majority of
reported 'bugs' on the ReiserFS these days are in fact hardware flaws,
not ReiserFS failings. I should point out that I have no vested
interest in defending ReiserFS and not particular love for it (and in
fact I'm pulling for COSS or another cyclical dedicated Squid store for
the long run solution to the Squid disk i/o issue), but it makes no
sense to spread FUD about a quite solid FS. If you prefer BSD and want
to encourage it's use, that's fine, but spreading misinformation is
counter-productive for everyone.
--
Joe Cooper <joe@swelltech.com>
Affordable Web Caching Proxy Appliances
http://www.swelltech.com
Received on Thu May 10 2001 - 04:04:45 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:59:54 MST