geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Matt Hogstrom <m...@hogstrom.org>
Subject Re: daytrader benchmark results
Date Thu, 16 Mar 2006 03:59:18 GMT


Maxim Berkultsev wrote:
> Hi Matt,
> 
> thanks a lot for the reply. A few more questions...
> 
> 
>>The Y-Axis represents transactions per second.
> 
> 
> What is this value - is it a general magnitude like time or is it related to
> the type of request which is under study, (i.e. number of JDBCRead
> transactions per second)?
> 
> 
>>All the TPS metrics were gathered from the client.  The goal was to get the
> 
> server to 100% CPU utilization.
> 
> Before your reply I suppose that Web Server Manager portlet was used to
> perform monitoring, please see the post
> 
> http://www.mail-archive.com/dev@geronimo.apache.org/msg18979.html
> 

No, the results are based on external requests per second by the client.  Not by 
the servlet mentioned in the article.

> So the question is - do you use any additional client application (not
> included in Daytrader), which provides workload for Geronimo?
> 

I use a load driver call WebSphere Studio Simulator.  This is a tool  that I 
have had setup on my test environment for some time.  At some point I'd like to 
move to JMeter or some other load driver but I found the Java based load 
generators take too much compute resource so right now I'm using a C based one.

> Thank you.
> 
> --
> Best regards,
> Maxim Berkultsev, Intel Middleware Products Division
> 
> Matt Hogstrom wrote:
> 
>>Hi Maxim,
>>
>>Comments inline...
>>
>>Maxim Berkultsev wrote:
>>
>>>Hello all,
>>>
>>>I've looked through Daytrader workload results and analysis for Geronimo
> 
> published at
> 
>>>http://www-1.ibm.com/support/docview.wss?uid=swg27006724
>>>
>>>For the bar diagrams, which show Geronimo and 'target' server indicators,
> 
> immediate questions are:
> 
>>>- what is the measure unit for values in Y-axis?
>>
>>The Y-Axis represents transactions per second.
>>
>>
>>>- how are the values for 'target' performance received?
>>
>>One problem with performance measurments is that they typically incite many
> 
> people to claim victory or cry foul depending on the results.  This is also
> 
>>complicated by the fact that many commercial products have a no benchmark
> 
> publish clause in their licenses.  In order to have something to compare
> 
>>Geronimo to as well as avoid the inevitable fallout of naming a comparison
> 
> point I did a series of measurments on other Open Source and commercial
> 
>>AppServer products.  I thook these results and created the competitive
> 
> target metric.
> 
>>The competitive target metric is simply my estimation of a respectable
> 
> performance measurement.  Meaning, as we achieve that metric one should be
> happy
> 
>>with the results.
>>
>>Note that these numbers are pretty old.  I have a new set of data that I've
> 
> been saying will be out for a while.  I guess its time to actually make that
> happen.
> 
>>So, for purposes of consumption I would wait a few days for the new report
> 
> to surface.  We'll post it on our site.
> 
>>
>>>Also, could someone clarify if the performance data was collected on a
> 
> client or a server side?
> 
>>All the TPS metrics were gathered from the client.  The goal was to get the
> 
> server to 100% CPU utilization.
> 
>>Also, if you guys are willing we could use some of the newer Potomac
> 
> 3.6Ghzchips:)
> 

Mime
View raw message