Jason Thompson wrote:
> 1. What is a recommended cache size to gain best performance? is
> there a size where you stop getting any performance gain or is it
> simply the bigger the better?
It all depends on your funds. Larger caches requires more disk and
memory, and percieved speed very much depends in connection bandwidth.
> 2. How much physical memory should the cache have in the machine,
> and how much should squid have allocated in the squid.conf file?
In your setup cache_mem should probably be 2 MB.
Also, remember to limit (or even disable) memory pools.
memory_pools_limit 1 MB
or
memory_pools off
> 3. I want to get the cache running as fast as possible for the
> following system below:
> ISDN Link to the net at 64Kbs
> Approx 100 Clients
> Server Currently: P166, 32MB RAM, 2GB HD
> OS: Redhat 6.0
With only 64Kbps link it is hard to configure a cache to not be able to
handle it.
> Could anyone offer me some tips for tuning the cache. It would
> be best if the cache made as little access to the Internet as
> possible. i.e not bother checking expire dates for cache objects. As
> long as users can manually refresh the pages using the web browser.
Problem is that is hard to distinguish between public and private
content when you begin to bend expiry rules. Some servers return private
content with no indication that it is private except that the page
contains no expiry information.
If you want to experiment with this, see refresh_pattern and
offline_mode settings, but be prepared to see problems with private
content like web based email access.
-- Henrik Nordstrom Squid hackerReceived on Sat Nov 20 1999 - 06:44:46 MST
This archive was generated by hypermail pre-2.1.9 : Wed Apr 09 2008 - 11:57:32 MDT