Amy (?),
you're over-simplifying somewhat.
the whole idea of a bakeoff is for a "realistic" test of various vendors'
caches.
in order for any test to be "realistic", it needs to mimic a real-world
environment.
indeed, for the second BakeOff, the concept of introducing "real" network
delay (in the form of a software solution under FreeBSD) adds to the
"realistic"ness.
unfortunately, we don't believe that the 'unif' object distribution model
is realistic.
in fact, Alex and Duane have said as much .. that it serves the purpose of
'preventing artificially high memory hit ratios'.
all we're pointing out is that it _isn't_ realistic - and giving some
pointers to what we believe _are_ realistic.
indeed, yes. we've spent considerable effort researching what real object
distribution patterns are.
NLANR themselves have an awful lot of real-world logs (*) to test the
validity of zipf, unif or any other distribution model.
[(*) see ftp://ftp.ircache.net/Traces/]
to my knowledge, end-users don't actually use NLANR's caches directly, and
this may cause some skew in the data, but i'm sure both (a) we and (b)
other vendors are more-than-happy to provide log files from real customers
that can help generate a "realistic" object distribution pattern.
lets keep the discussion factual rather than opinion based, ok?
cheers,
lincoln.
At 18:42 30/11/99 -0800, ipv6 wrote:
>For some reason, I have a feeling that
>Cisco's cache can handle Zipf model
>well, but not so well if the Unifom
>model is used...
>
>I tought Cisco can use the strength of the
>Wisconsin's design to handle the Unifom
>model well.
--
Lincoln Dale Content Switching
ltd@cisco.com Cisco Systems Inc. | |
|| ||
+1 (408) 525-1274 bldg M, 170 West Tasman |||| ||||
+61 (3) 9269-4200 << San Jose CA 95134 ..:||||||:..:||||||:..
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:09 MDT