Re: Cache-DIGEST false hits

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Wed, 19 Aug 1998 02:41:27 +0200

I think I have to agree with Serge Krashakov on this. His figures
indicates that there is a large number of false hits.

There is a large difference between CACHE_DIGEST_HITs on the first
cache,
compared to the number of cache hits on the second cache when counting
requests from the first cache.

And if this is a sibling relationship, then there for sure is a
high false-hit ratio no matter if cache-digests is involved or not
(which they are), since the number of misses is far greater than
the number of hits.

Only thing that troubles me is that the sums does not add up between
the two caches, which may indicate that the logs are not from the same
period, but it it is a sibling relation then this does not matter to
much as only requests from the first cache is measured in the second
caches report.

/Henrik

Alex Rousskov wrote:
>
> Hi Serge,
>
> I am not sure why you think that most of CACHE_DIGEST_HIT are false
> hits. The tables you have posted do not contain explicit information about
> false hits as far as I can tell. Does Calamaris support CACHE_DIGEST tags?
> Please explain what makes you think that the number of false hits is too
> large.
>
> Thanks,
>
> Alex.
>

On Tue, 18 Aug 1998, Serge Krashakov wrote:

> # Outgoing requests by destination
> neighbor type request % kByte % sec kB/sec
> second-cache 3126 2.66 8328 0.87 0 0.24
> CACHE_DIGEST_HIT 3053 2.60 7838 0.82 0 0.23
> SIBLING_HIT 73 0.06 490 0.05 0 0.50

> # Incoming TCP-requests by status
> status request % kByte % sec kB/sec
> --------------------------------- -------- ------ --------- ------ ---- -------
> HIT 1143 36.33 2250 20.35 0 4.98
> MISS 1976 62.81 6355 57.48 20 0.16
> Sum 3146 11056 13 0.28
Received on Tue Aug 18 1998 - 18:13:31 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:41:39 MST