DNS Performance Numbers
kreios at gmail.com
Wed Oct 25 15:56:44 UTC 2006
On Oct 25, 2006, at 2:35 AM, Divacky Roman wrote:
> On Tue, Oct 24, 2006 at 09:59:57PM -0500, kreios at gmail.com wrote:
>> I am running some performance tests on named to see how it performs
>> with different configurations on FreeBSD and figured I would share
>> first results. The first tests are for serving up static data.
>> Supermicro PDSMi Motherboard
>> 1G Memory
>> Intel Pentium D CPU 3.40GHz
>> Intel Gigibit NIC
>> Bind 9.2.3
>> OS UP UP+P MP MP+P MP+TP MP+TT MP
>> FreeBSD 4.11 28455 28370 28976 X X X
>> X X
>> FreeBSD 6.1 29074 34260 34635 35730 17846 38780 19776
>> FreeBSD Stable 30190 34707 33294 36651 18893 39374 19449
>> FreeBSD Current 30707 34029 32300 33689 15535 40554 13886
>> Ubuntu 6.06 X X X X X 37294
>> X X
> I see regression between -current and -stable. are you sure you
> tested without
> any debuging stuff? some performance speedups went in in 7-current
I would attribute the regression to my testing setup. Running a couple
of the tests again, I get numbers more in line with stable and release.
In addition, I was probably not as careful with current as we will not
be running it in production. I will see if I can find the time to run
the numbers again and update the table.
FYI, this first round of testing was to answer the question, does using
threaded bind help performance on FreeBSD? In this minimal test case, it
does. With the above info, I can now look into how a bind vs threaded
bind performs in different test cases. This also gives me information on
what I will gain/lose when deciding on various administration issues.
> also - do you use the same config everywhere? -current GENERIC
> doesnt have COMPAT_43
> for example which miht affect performance (additional locking) etc.
I modified the kernel configuration files that where included with the
distribution. For release, stable and current, I created four kernels;
uni-processor, uni-processor with polling, multi-processor, multi-
processor with polling. The OS was as default as possible. For this
test, I didn't want to make many changes.
More information about the freebsd-performance