users@grizzly.java.net

Re: why grizzly performance low?

From: Barry van Someren <barry_at_coffeesprout.com>
Date: Tue, 21 Jul 2009 13:09:14 +0200

Hi,

Small addition (sorry to reply to my own email)
The root cause of my hang was running out of FileDescriptors.

Disregard the PS section of my last email.

Barry

On Tue, Jul 21, 2009 at 1:05 PM, Barry van
Someren<barry_at_coffeesprout.com> wrote:
> Hi,
>
> Could it be that you are running into a bandwidth constriction?
> I've just gotten done running some static file performance tests on
> Glassfish V3 and here are some of my results:
>
> Lifting the server siege...      done.
> Transactions:                  15008 hits
> Availability:                 100.00 %
> Elapsed time:                  46.18 secs
> Data transferred:            5166.84 MB
> Response time:                  0.38 secs
> Transaction rate:             324.99 trans/sec
> Throughput:                   111.88 MB/sec
> Concurrency:                  124.47
> Successful transactions:       15008
> Failed transactions:               0
> Longest transaction:            3.41
> Shortest transaction:           0.01
>
> The files range from 1024 bytes to 1 Mb (x*x*1024) and the requests
> take a random sample.
>
> At double (250 concurrent  requests per sec) we get:
>
> Lifting the server siege...      done.
> Transactions:                  10417 hits
> Availability:                 100.00 %
> Elapsed time:                  32.27 secs
> Data transferred:            3593.03 MB
> Response time:                  0.77 secs
> Transaction rate:             322.81 trans/sec
> Throughput:                   111.34 MB/sec
> Concurrency:                  247.03
> Successful transactions:       10417
> Failed transactions:               0
> Longest transaction:            0.97
> Shortest transaction:           0.01
>
> As you can see we run out of network connectivity way before taxing
> the CPU (it was pegged at 80% of 1 core on a dual Intel Xeon 5530
> machine)
> Also, please include the correct data.
> PS: For some odd reason, if I start a siege with more than 400
> requests Glassfish just hangs, but that's another issue, probably
> fixed already as it's an old version.
>
> Regards,
>
> Barry
>
>
>
> On Tue, Jul 21, 2009 at 10:45 AM, Paul Sandoz<Paul.Sandoz_at_sun.com> wrote:
>>
>> On Jul 21, 2009, at 5:29 AM, wy_0844_at_sina.com wrote:
>>
>> jsp submit result:
>>     Server                TPS
>> JBoss
>> 263
>> GlassFish
>> 3710
>> Geronimo
>> 3278
>> Tomcat
>> 3443
>>
>>
>> huge datas result:
>>  Server                TPS
>> JBoss
>> 263
>> GlassFish
>> 3710
>> Geronimo
>> 3278
>> Tomcat
>> 3443
>>
>>
>> why glassfish have low performance in huge datas test?
>>
>> Did you incorrectly copy the results from the "jsp submit result" to the
>> "huge datas result" ?
>> Paul.
>>
>> we use jprofile find reason,watch jvm running time, cup call time,blocking
>> time,waiting time,threads status,but not find reason.
>> we watch glassfish src, it use NoParsingAlgorithm , is it low performance
>> reason?
>> other problem why a http call use SSLWorkerThread class?
>>
>> could you help me,hope you email.
>>
>>
>>
>> Thanks.
>>
>> Best Regards.
>>
>
>
>
> --
> Barry van Someren
> ---------------------------------------
> LinkedIn: http://www.linkedin.com/in/barryvansomeren
> Skype: BvsomerenSprout
> Blog: http://blog.bvansomeren.com
> KvK: 27317624
> irc: BarryNL @ FreeNode
>



-- 
Barry van Someren
---------------------------------------
LinkedIn: http://www.linkedin.com/in/barryvansomeren
Skype: BvsomerenSprout
Blog: http://blog.bvansomeren.com
KvK: 27317624
irc: BarryNL @ FreeNode