Re: [squid-users] cache digest based sibling selection

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Fri, 23 Jan 2004 16:41:05 +0100 (CET)

On Thu, 22 Jan 2004, Stuart Smith wrote:

> Can someone please explain to me the mechanism squid uses to select
> a sibling to fulfil a request if more than one sibling has the object,
> and I am using cache digests ?

In case of ICP the first one that answers.

> The faq says that it will use NetDB to select a cache.

Not if you use ICP.

If your are not using ICP then I am not entirely sure, but it is possible
NetDB may be used in such case (provided you use cachedigests, and there
is a digest hit).

> assuming I used cache digests, AND specified no-netdb-exchange, how would
> it make its decision ?

>
> Also, assuming no sibling has the object, it would seem beneficial in some
> circumstances to still ICP or HTCP check all siblings, since:
>
> 1) the cache digest could be nearly an hour old
> 2) the lossiness of the digest could have created a false miss.
>
> for example, if internet bandwidth is particularly limited and/or expensive.
>
> It is unclear how/if ICP/HTCP works with cache digests on.
>
> It seems to hint that digests are an alternative to ICP/HTCP without
> specifically saying that they are mutually exclusive ?
>
> Stuart C. Smith
> Technical Architect
> Venation
> www.venation.com
>
Received on Fri Jan 23 2004 - 08:41:12 MST

This archive was generated by hypermail pre-2.1.9 : Sun Feb 01 2004 - 12:00:08 MST