perfsonar development work

[Jason Zurawski <>]

Nina Jeliazkova wrote:
> Hi,
>
> Slawomir Trzaszczka wrote:
> > On Fri, 2009-11-27 at 11:06 +0200, Nina Jeliazkova wrote:
> >   
> > > Slawomir,  All,
> > >
> > > Very interesting comparison. 
> > >
> > > It would be also useful to test sending requests ( queries at least)
> > > in parallel, if possible, since I imagine the normal  usage pattern is
> > > handling several requests at once.
> > >
> > >     
> >
> > Parallel or in some sequence (one thread) ? Parallel tests are hard to
> > interpret (threading, concurrency access) and main problem with
> > repeatable.
> >
> >   
> I am just curious whether it is possible to mimic a real use case , of 
> course it can be hard to measure.


I agree that this is something worth trying for.  In our experience we 
see the biggest performance hit during concurrent operations:

 1) large registration coupled with complex queries
 2) service cleaning coupled with registration/queries
 3) service summarization coupled with registration/queries

-jason