a very quick perf comparison of 9.1.0b3, b2, and 8.2.2pl5
Rick Jones
raj at cup.hp.com
Tue Jan 9 19:20:42 UTC 2001
Here is a very quick perf comparison between three revisions of bind,
all compiled "stock" as the source shipped - in particular with -g. The
system is an hp visualize j5000 with two 440 MHz PA-8500 CPUs. In the
case of BIND 8 one instance of named was running and it was run out to
saturation of one CPU, with theother CPU basically idle. In the case of
BIND 9, named was explicitly told that there were two CPUs and they were
both run out to rather close to saturation with virtually all the time
in user mode.
Named Perf on Two CPU J5000
Stock compilation - -g
8.2.2pl5 9.1.0b2 9.1.0b3
Threads of Load Ops Ops Ops
1 1812 864 1143
2 3634 1549 2051
3 3668 1674 2246
4 3665 1845 2549
5 3670 1995 2846
6 3666 2066 2954
7 3666 2127 3040
8 3662 2159 3072
9 3662 2133 3105
10 3658 2177 3162
11 3662 2168 3201
12 3662 2210 3198
13 3663 2203 3206
14 3661 2239 3238
15 3664 2213 3213
16 3660 2251 3271
b3 does indeed have a nice big boost in perf and increases single
instance named9 scaling from 0.62X to 0.89X of named8 on a two CPU
system. I hope to have some time to coerce the bits into compiling with
something other than -g and recreate more of the data from:
ftp://ftp.cup.hp.com/dist/networking/briefs/dns_server_results.txt
rick jones
--
ftp://ftp.cup.hp.com/dist/networking/misc/rachel/
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to email, OR post, but please do NOT do BOTH...
my email address is raj in the cup.hp.com domain...
More information about the bind-workers
mailing list