BIND9 performance issues with SMP

[JINMEI Tatuya /神明達哉]

>>>>> On Tue, 21 Dec 2004 17:17:32 -0700, 
>>>>> Scott Long <scottl at freebsd.org> said:

>> C. (for comparison) SuSE Linux (kernel 2.6.4, glibc 2.3.3) on the
>> same box I used with experiment B
>> 
>> threads  BIND  BIND++
>> 0        16117
>> 1        13707  17835
>> 2        16493  26946
>> 3        16478  32688
>> 4        14517  36090
>> 
>> While "pure BIND9" does not provide better performance with multiple
>> CPUs either (and the optimizations in BIND++ are equally effective),
>> the penalty with multiple threads is much smaller.  I guess this is
>> because Linux handles lock contentions much better than FreeBSD.
>> 

> Do you have any comparisons to NetBSD or Solaris?  Comparing to Linux
> often results in comparing apples to oranges since there is
> long-standing suspicion that Linux cuts corners where BSD does not.

I've never done this type of test for NetBSD, since as far as I know
NetBSD is not very SMP-aware (does this change in, e.g., NetBSD 2.0?).

I've checked Solaris with similar tests, but I could only use
a 2-processor sparc box.  So, the results would not be very
informative.  FWIW, however, Solaris performed quite well with 2
processors.

> Also, would you be able to re-run your tests using the THR thread
> package?

If I have another chance and test environments (I've lost the access
to the test environments).

					JINMEI, Tatuya
					Communication Platform Lab.
					Corporate R&D Center, Toshiba Corp.
					jinmei at isl.rdc.toshiba.co.jp