Re: [squid-users] cache_dir slection criteria

> On Wed, 24 Nov 2004, Hendrik Voigtlaender wrote:
> >I think I could solve my problem by using round-robin, which is pretty
> >much self explaining. One question remains: What is squid's opinion of
> >"least-load" ?

On 09.12 00:07, Henrik Nordstrom wrote:
> Depends on the cache_dir type.
>
> For ufs it is the number of currently open files (squid-2.5.STABLE7,
> earlier versions consider all equal load).
>
> Fur aufs/diskd it is the number of pending operations within the
> cache_dir.

That results into different usage of drives of different speeds/sizes.
the slower is the drive, or the bigger cache size is, the more requests are
usually waiting on it.

That would result into situation, where one drive is equally loaded than
x-times bigger and x-times faster one. Good that new bigger drives use to be
equivalently faster.

For different relative performance of drives, it may happen that relatively
slower drive (smaller and much slower, or bigger and not much faster) will
contain much older data, which will result in unoptimal disk use, but it is
probably not in our powers to change it.

I was thinking about making efficiency of slower harddrives better by
disabling storing of small files on them (but the min-size option for
cache_dir does not exist), or setting max-size to cache_dir on faster disks,
both to prefer storing of bigger files to those slower disks.
(I expect that small files cause higher load on disks).

This can be somehow limited by having big cache_mem setting with small
maximum_object_size_in_memory (I currently have 256M / 256K)

However it's just guessing, it is hard to know how will disk configuration
behave without really trying it.

-- 
Matus UHLAR - fantomas, uhlar@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.
Despite the cost of living, have you noticed how popular it remains?