Re: Disk I/O

On Friday 11 October 2002 16.09, Денис Овсиенко wrote:

> I absolutely agree on index things (see above), but it naturally
> comes to storing objects' body in the same database to not
> duplicate file sizes and to use the same way to extract them.
>
> P.S. It's just an idea, that seems useful to me and I wished to
> donate it here.

You are more than welcome to help and try out the idea and prove Joe
and others wrong here.

However, one immediate problem is the current code requirement of a
in-core cache index (there must be no delays in index lookups), but
the refactoring of storeGetPublicByRequest() Robert is doing aims at
making of it possible to have a "on-disk" index stored elsewhere than
in the memory of the Squid process. The Squid development team has
the long-term goal that a "object store" should be able to choose
freely how it implements both storage and indexing.

Regards
Henrik
Received on Fri Oct 11 2002 - 10:13:26 MDT