Re: [squid-users] About Cache Digest

On Mon, 2008-04-07 at 10:43 +0200, Henrik Nordstrom wrote:
> mån 2008-04-07 klockan 11:44 +0800 skrev Zhou, Bo(Bram):
>
> > 1. How often local host fetch the digest from peers?
>
> Each digest_rebuild_period as set by the peer.
>
> > Any timer controls or
> > just checks if the cache digests are expired? Some hard-coded timers,
> > PeerDigestReqMinGap and GlobDigestReqMinGap in peer_digest.c, are defined in
> > the source codes, what's purpose of these timers? When will these timers be
> > started and stopped?
>
> These are safeguards to make sure digest fetching doesn't suck up too
> much bandwidth, in case the peer(s) is badly configured.
>
> > 2. For digest_rebuild_chunk_percentage, if I define it 10%, does that mean
> > Squid will add more 10% cache to digest for each digest_rebuild_period
> > interval or within digest_rebuild_period interval but finish processing
> > other user requests?
>
> It's how much of the cache Squid should digest at a time when rebuilding
> the cache digest. No requests is processed while this is done, only very
> briefly between chunks.
>
> Hmm.. not even sure why this directive exists. I think it should scan
> one bucket at a time.. digest rebuilding is not a very high priority
> item and blocking request forwarding for this is not good...
>
> Alex: do you remember why this directive was added?

IIRC, one-bucket-at-a-time was too slow in environments that cared more
about hits than optimizing miss performance. Folks wanted a knob to
control the trade-off.

HTH,

Alex.
Received on Tue Apr 08 2008 - 08:57:21 MDT