Re: [squid-users] Performance tips for accelerator setup

Markus Meyer wrote:
> Markus Meyer schrieb:
>> What is your reason for using heap LRU?
>> I've heard/read this before that it performs better than LRU. So it's on
>> my todo-list. But I would like to understand better why heap LRU should
>> be better.

plain lru is just a list walk and removal. It risks falsely removing old
objects that are infrequently used but very cachable.

heap seems to have a faster time locating just the objects that are not
going to be needed again. "heap LRU" is a hybrid that uses the fast
access approach and does LRU removal on them as well to be extra safe.

The result is slightly better hit ratios. Growing better as cache size
and traffic load grows.

>
> One idea I have is to use two different replacement policies for memory
> and disks:
>
> mem - Heap GSDF: fast and small objects preferred. Better to have more
> small objects in memory than to read them from disk.
>
> disk - Heap LFUDA: fast and more popular pictures, more "hot" content,
> is preferred.
>
> But that's just a wild guess.

Close.

I'd add a COSS directory for best access to the small items going to disk.

Also, you may want to use CARP from the parents to siblings. That will
restrict the objects each sibling is caching to the URLs CARP assigns
it. Less duplicated objects leading to a longer caching time window and
more hits overall.

If you are using ICP between the siblings, try HTCP for better peer HITS.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE7 or 3.0.STABLE21
   Current Beta Squid 3.1.0.15