Open MPI logo

MTT Devel Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all MTT Users mailing list

From: Ethan A Mallove - Sun Microsystems (Ethan.Mallove_at_[hidden])
Date: 2006-09-05 10:19:34


Josh Hursey wrote On 09/05/06 10:04,:
> On Sep 5, 2006, at 8:43 AM, Jeff Squyres wrote:
>
>
>>IU / HLRS -- How's it going? I see a bunch of submits from you
>>guys in the
>>database, so I assume things are running smoothly. But let me know
>>one way
>>or another.
>
>
> IU is setup and running on our Odin cluster.
>
> Nightly we run 8 nodes (16 procs):
> trunk,v1.2 (gcc)
> IBM,Trivial,Intel,IMB
> Currently just using tcp,self (due to some cluster mgmt issues)
>
> Weekly we run 32 nodes (64 procs):
> Same configuration as the nightly
>
> We are still doing some local debugging (not MTT issues really, just
> cluster stuff) so commits to the MTT DB may come a bit more
> frequently now then they would normally.
>
> We may have some other clusters coming online in the future, but we
> can't give specifics at the moment.
>
> I have been putting any bugs I see into Trac as they popup (sorry for
> duplicates).
>
> It seems that the DB 'cleans' itself every so often. Meaning that I
> notice that runs from a week or so ago are no longer in the summary
> page. Is this true? How often does it do this?
>

summary.php only shows results from yesterday through "now". The data submitted
before that is not deleted, just not shown. We have a drill-down tool in the
works which will allow for more flexibility wrt a bunch of parameters.

> Cheers for all the hard work. This is looking really nice and useful.
>

That's what we like to hear - thanks!

-Ethan

> -- Josh
>
>
>>We have a bunch of improvements waiting in the wings:
>>
>>1. The version that you are using makes N connections to the IU web
>>server
>>to submit N results (i.e., one connection per result). This is not a
>>problem for many people, but I kept getting cutoff by the Cisco web
>>proxies
>>because they thought MTT was a worm. So I implemented "bulk submit",
>>meaning that results are gathered up (essentially a test run
>>section at a
>>time) and submitted in one connection to the database. Not only
>>does this
>>make MTT not look like a worm, it's a bit more efficient (which is
>>not a
>>necessarily primary concern, but it is a Good Thing).
>>
>>We need to coordinate when to deploy this to you because it affects
>>both the
>>"mtt" client and the PHP at the URL that you are submitting to. More
>>details to come on this. I realize that you guys apparently don't
>>*need*
>>this, but a) I do, and b) others might when we deploy MTT to the
>>entire
>>group.
>>
>>2. Various bug fixes have been committed to the release branch over
>>the last
>>week as a direct result of your feedback. Keep the comments coming!
>>
>>3. Sun is actively working on reporting the data in spiffy web
>>pages. Stay
>>tuned on this front.
>>
>>--
>>Jeff Squyres
>>Server Virtualization Business Unit
>>Cisco Systems
>>_______________________________________________
>>mtt-users mailing list
>>mtt-users_at_[hidden]
>>http://www.open-mpi.org/mailman/listinfo.cgi/mtt-users
>
>
> _______________________________________________
> mtt-users mailing list
> mtt-users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/mtt-users