On Wed, 2 Jul 1997, Dave Zarzycki wrote:
>
> Well in that case, can multiple squid processes share a cache directory?
> I would imagine there would be problems with locks on files, etc. Could
> this be solved by a master/slave relationship, or a separate program for
> managing the cache? This would have other applications too, such as
> several machines hooked up to a large cache via SSA, or FireWire.
But are you really finding that Squid is becoming CPU-bound, rather than
memory or disk I/O bound? My cache has been up for about two days,
averaging 1,110.8 connections per hour (52,886 connections), and although
this is probably pretty small when it comes to caches, the CPU Usage
figure is sitting at about 1%. My average select loop time is 93.341 ms.
> I suppose that I could customize my proxy auto config file and squid.conf
> file to specify something such as the following for a MP machine:
>
> 2 Processors
> 2 squid processes
> One ip_alias entry for a card to make life pretty
I guess the ultimate key question is does your operating system support
multiple processors? If it doesn't, all this is just moot.
I don't know, though, I really get the sense that multiprocessing is just
a big marketing gimmick that has utility only if you're running piggy,
inefficient operating systems like Windows NT, or certain specialized
applications. For example, a single processor BSD/OS machine posted
numbers comparable to Windows NT systems on dual-processor boxes.
See <http://www.techweb.com/se/directlink.cgi?CWK19970505S0103> -- look
for the paragraph sarting with "Making matters even more interesting".
-Mike Pelletier.
Received on Thu Jul 03 1997 - 07:07:34 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:35:40 MST