Yes, very awesome to have and thanks for fixing that Julien!
I have created a demo (using the bridge) leveraging the available JMX
data. I have also created a "live view" feature which allows for the
charts to refresh in real time at specified intervals (using polling)
- I was going to use ajax push, but I found a bug in the bridge that
does not return the response.
So for a very simple test, I used Jmeter and created a 20 user/thread
load with 5 loop count, and then added every default portlet to a
single page - and watched my macbook fall to its knees ;) but it is
cool to see the statistics that it generates.
The code is pretty much ready at this point, I am just making things
look attractive and also using it to test questionable RF components
for EPP support.
Here is a preview
http://img4.yfrog.com/i/picture5eh.png/
On Nov 16, 2009, at 9:00 AM, Julien Viet wrote:
On Nov 16, 2009, at 2:50 PM, Prabhat Jha wrote:
> Are these stats on by default?
yes
> What is the overhead of getting these metrics?
The collecting of data does not take much overheard, I have never
spotted them in my JProfiler so far.
The processing of collected data must not be free, but I would not
expect to have a sampling with a one second period to introduce much
bias.
> BTW: It's awesome data to have.
thanks :-)
>
> ----- Original Message -----
> From: "Julien Viet" <julien(a)julienviet.com>
> To: gatein-dev(a)lists.jboss.org
> Cc: "Wesley Hales" <whales(a)redhat.com>, "Marek Posolda"
<mposolda(a)redhat.com
> >
> Sent: Monday, November 16, 2009 4:28:20 AM GMT -06:00 US/Canada
> Central
> Subject: [gatein-dev] Fix in portal list
>
> I want to bring to you that useful following statistic available via
> JMX, specially in JBoss where you can use the JMX console.
>
> Wesley has been using it for his bridge demo (and I fixed the bug
> that
> was present Wes :-) ).
>
> I believe also that Marek could collect information during load
> testing and create graphs that would be correlated with the
> performances graphs. (for instance , showing the portal throughput
> and
> the average response time on the same graph, or showing the usage of
> the various caches correlated with the througput, I guess you get the
> idea).
>
> With the JMX console, the statistics are interesting but it remains
> limited, it becomes very useful only if we can exploit the data via a
> more powerful tool with charting and gauges.
>
> 1/ service=statistic,type=portal,view=portal
>
> - gives you the list of portal ids
> - for a given portal id you can get info like:
> 1/ average execution time
> 2/ minimum execution time
> 3/ maximum execution time
> 4/ execution count (can be graphed)
> 5/ throughput, i.e number of request per second (can be graphed)
>
> 2/ service=statistic,type=template,view=portal
>
> - list of the 10 slowest templates
> - list of the 10 fastest templates
> - list of the 10 most used templates
> - list of all template ids
> - for each template it is possible to have info about:
> 1/ average time
> 2/ min time
> 3/ max time
> 4/ execution count
>
> 3/ service=SessionRegistry
>
> it gives the list of the number of active JCR session in the portal
> which can be graphed as it gives a good indicator of the load on the
> server
>
> 4/ each cache is registered under the key
> name="$CACHE_ID",service=cache which gives the usual information
>
> there are other few mbeans available but they are less interesting,
> so
> I won't talk much about it.
>
> cheers
>
> Julien
> _______________________________________________
> gatein-dev mailing list
> gatein-dev(a)lists.jboss.org
>
https://lists.jboss.org/mailman/listinfo/gatein-dev
Wesley Hales
JBoss Portal Developer
whales(a)redhat.com