Stephen Smalley wrote:
>In the past, we haven't found kernel compile benchmark to be very
>revealing for SELinux performance analysis. dbench results would be of
>interest. More generally, you might want to repeat the tests done for
>the AVC RCU work, see
>
>http://marc.theaimsgroup.com/?l=linux-kernel&m=110054824004161&w=2
>
>
Hmm. What do you mean that kernel compile is "not very revealing" for
SELinux performance analysis?
To be sure, kernel compile (khernelstone :) is a macrobenchmark, and
will not give you high-resolution data on the cost imposed on individual
operations. For microbenchmarks, I prefer lmbench.
For macrobenchmarks, you want something that does lots of stuff, so that
the run times are long enough to measure, and does the right mix of
stuff, so that the workload is representative for "typical" workloads.
Of course, "typical" is in the eye of the beholder.
At Immunix, we tend to use 2 macrobenchmarks: khernelstone, and
Webstone. Kernel compile does lots of everything (CPU, disk, and memory)
presenting a heavy workload, and thus *should* make a good benchmark,
unless the workload mix is somehow wrong.
Stephen, what do you find wrong about the workload mix in kernel
compiles? What would you suggest instead?
Thanks,
Crispin
--
Crispin Cowan, Ph.D. http://immunix.com/~crispin/
CTO, Immunix http://immunix.com
This archive was generated by hypermail 2.1.3 : Wed Mar 09 2005 - 15:19:21 PST