Re: BIND9 performance issues with SMP

JINMEI Tatuya / $B?_at_L_at_C#:H wrote:
>On Tue, 21 Dec 2004 17:17:32 -0700, Scott Long <scottl_at_freebsd.org> said:
[ ... ]
>>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?).

Indeed, from http://www.netbsd.org/Releases/formal-2.0/NetBSD-2.0.html

"The addition of a native threads implementation for all platforms and 
symmetrical multiprocessing (SMP) on i386 and other popular platforms were 
long-standing goals for NetBSD 2.0. Both of these goals have now been met—SMP 
support has been added for i386, SPARC, and PowerPC, the SMP support on Alpha 
and VAX has been improved, and the new port to the 64-bit AMD/Opteron also 
supports SMP."

> 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.

Solaris probably has the best SMP/threading implementation available today, 
and the SPARCv8 and v9 architectures were highly oriented towards supporting 
parallel execution and dealing with SMP cache coherency issues.  Solaris on 
SPARC scales up with more CPUs added against workload very well, the only real 
problem the platform had is that individual SPARC CPU's aren't especially fast 
to begin with.

Solaris is also using streams rather than a classic BSD TCP/IP network stack, 
although FreeBSD itself is going beyond classic TCP/IP stuff via Netgraph...

-- 
-Chuck