To answer my own question: No
Last week I did a quick comparison of the speed of Squid.NOVM.9 and
Squid.NOVM.20, which seemed to show the old squid was faster, but
as Alex Rousskov (and others) pointed out:
>1) Do not use mean for average response time; it is virtually meaningless.
>Use median instead.
>
>2) A distribution of response time rather than single median is better, of
>course.
>
>3) Measure it for several days, not 10,000 transactions.
So I did some longer term measurements, using medians instead of average
access time, and there was no appreciable difference in speed. Here
are the results for three days with each version of Squid:
TCP_HIT #Requests TCP_MISS #Requests
Squid.NOVM.9
314 msecs 194,602 1391 msecs 193,178
455 msecs 251,170 1958 msecs 194,129
374 msecs 236,287 1648 msecs 178,613
Squid.NOVM.20
352 msecs 174,072 1642 msecs 165,815
382 msecs 227,845 1744 msecs 190,925
327 msecs 224,076 1475 msecs 179,361
-- Earl Fogel Computing Services phone: (306) 966-4861 University of Saskatchewan email: earl.fogel@usask.caReceived on Fri Feb 13 1998 - 17:03:36 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:38:53 MST