Do not test at localhost... NEVER.

Just a bit of "luck" and you can get completly different results...
One CPU ( even with 2-4 cores ), that's still one CPU, running a
webserver with threads and benchmark, you will get "weird" results...

Typical problem with different multiple applications. That's fight for
CPU and RAM resourcess... NEVER do benchmarks on localhost...


Btw ( send your email to mailing list too )

Pozdrawiam
Jędrzej Nowak



2009/7/28 Eric Lapouyade <[email protected]>:
> Hi,
>
> Yes, I am testing on localhost, I do not have 2 servers side by side.
> It does not test the tcp/ip stack, but gives a good idea what could be
> performance for distant connexion.
>
> Having only 500 Req/s for keep-alive connections instead of 10 000 Req/s
> with no keep-alive, even on localhost, it seems to be a big bug.
> When I test with cherokee 0.6 and 0.7, the perf was good on localhost and
> keep-alive, since 0.8, I get 20x less perf... Looks like a bug or a
> configuration problem on my linux Ubuntu 9.04 (I install cherokee on 2 x
> Ubuntu 9.04, still the same perf ...)
>
> Eric
>
> 2009/7/28 Jędrzej Nowak <[email protected]>
>>
>> Eric, do you still benchmarking localhost ? Or even 2 servers on the
>> same machine ? You MUST benchmarking two independent machines with
>> "isolated" environment...
>>
>>
>> Pozdrawiam
>> Jędrzej Nowak
>>
>>
>>
>> On Tue, Jul 28, 2009 at 6:32 PM, pub crawler<[email protected]>
>> wrote:
>> > I am not a stickler about benchmarking - but environments are varied
>> > as well as tests. It's good to see people spending time to test and
>> > contribute their results - even if the Cherokee community doesn't like
>> > the results at first glance.
>> >
>> > Post as much as you can about the tests that have been ran -
>> > environmental details, sample files, testing tools, etc.  - so the
>> > rest of us can replicate your benchmarking.
>> >
>> > Needless to say Cherokee development is active and ongoing and any
>> > perfomance issues that might show up should be quickly improved upon.
>> >
>> > On Tue, Jul 28, 2009 at 12:27 PM, kevin beckford<[email protected]>
>> > wrote:
>> >>
>> >>
>> >> On Tue, Jul 28, 2009 at 8:44 AM, Pablo Hernan Saro
>> >> <[email protected]>
>> >> wrote:
>> >>>
>> >>> Hi Eric,
>> >>>
>> >>> First of all, please don't take it the wrong way, but... What is your
>> >>> experience in doing benchmarks?
>> >>
>> >> Nope.  Wrong question.
>> >> "The Fastest free Web Server out there!"
>> >>
>> >> is from the project homepage.  Nothing wrong with this, it's good advertising, and cherokee is a great project.  What we need is an updated benchmark if the above statement still is true, and if it's not, switch marketing to memory use and cherokee-admin.
>> >> After all, ideally the fastest free web server out there should show
>> >> such
>> >> speed in a variety of benchmarks no?  If not, then who are the master
>> >> benchmark people?
>> >> _______________________________________________
>> >> Cherokee mailing list
>> >> [email protected]
>> >> http://lists.octality.com/listinfo/cherokee
>> >>
>> >>
>> > _______________________________________________
>> > Cherokee mailing list
>> > [email protected]
>> > http://lists.octality.com/listinfo/cherokee
>> >
>> _______________________________________________
>> Cherokee mailing list
>> [email protected]
>> http://lists.octality.com/listinfo/cherokee
>
>
_______________________________________________
Cherokee mailing list
[email protected]
http://lists.octality.com/listinfo/cherokee

Reply via email to