Here is the current status of the work.
1) The IDL visualization problem with the data being written by efit is a very
obscure bug. It appears that some of the data is getting corrupted somehow
which is causing some of the computations done in the IDL animation code
to get into an infinite compute loop. I believe I will be able to work on this
without
the sc computers next week. In the mean time we do have a working demo which
is reading from the datafiles that don't have this problem.
2) I've added the checking for the status in the netlogger calls and re-establish
the connection to the NetLogger if the writes fail and re-attempt the writes.
3) Qian has installed the efit program on the sc cluster and it works fine!
I've set it up to run from my account for the demo. IMPORTANT>> Apparently,
the sc cluster is not recognizing the scdemo credentials. There is no map for the
scdemo credential to my account in the gridmap file. <<IMPORTANT
Mary,
I've put an scemo-idl.tgz in diesel.lbl.gov:~twf which has the NetLogger
instrumented controller. On natasha you will need to:
# cd /usr/local/fusionDemo/idl
# tar zxf scdemo-id.tgz
If you want to run the controller with the NetLogger you'll need to add a line to
# echo 'netlogger_server terra.mcs.anl.gov:14835' >>
/usr/local/fusionDemo/config/envsyms
-tom
Kate Keahey wrote:
> All,
>
> we had the last telecon related to the sc demo today, so I am posting a
> summary of issues since they will have to be resolved on this list now.
> Please let me know if I forgot about something (practically a given ;-) or
> correct me if I misrepresented something.
>
> Fist some loose ends pertaining to the demo:
>
> 1) Qian and Tom are trying to make the IDL visualization work with data
> that is actually produced by efit (as opposed to the slightly "fake"
> approach we are using now). Qian will get an account on the Argonne sc
> machines, and will try to resolve it. If necessary she can also reproduce
> the problem on Tom's workstation so this can be worked on even after the sc
> machines are shipped if necessary.
>
> 2) Tom will add a check for write status to the netlogger writes
>
> 3) Qian will install efit on the Argonne sc cluster so we have a backup if
> the network goes down.
>
> 4) Chris will make some animations of the visualization of the data
> produced by efit and they will be accessible via a webpage. They require no
> special hardware or software beyond what comes with Windows to run. Chris
> will also send some explanations and images to David to produce a handout
> for the visualization part. Kate and Mary will meet with David on Sunday
> and get updated on what Fusion scientists want from visualization, and the
> ground breaking achievements this collaboratory will produce so they can
> later convincingly pretend to know something about it.
>
> The place to change default settings for the netlogger, database and efit
> is /home/twf/fusion/fusionDemo/config/envsyms (Tom, if I ever ask you
> again, which no doubt will happen, feel free to bounce this email to me
> ;-). Or they can be changed at command line.
>
> We also discussed some things that we see as accomplishments/interesting
> things about this demo. Here is a summary. Feel free to add/delete/disagree.
>
> 1. development
> - development comprised globus extensions to MDSplus and a couple of 10
> line shell scripts
> - took ~ 2 weeks
> - by application scientists
> - expert in application-specific software but with no previous exposure to
> globus
>
> 2. what this software represents
> - a real first step towards creating a fusion grid; the software will no
> doubt evolve, but on the existing base (as opposed to being replaced)
> - the components developed to gridify efit will be reused to gridify other
> Fusion applications, efit was a proof of concept
> - this is not a throwaway prototype
>
> 3. value added
> - the most important thing added over what MDSplus already had was secure
> environment
>
> 4. what made all this possible
> - the Fusion community was ready for grids (had the foresight, approach,
> software, expertise, need...) so that the port to grids was an extension of
> an existing approach rather than a revolution
> - we have the right grid abstractions to do this kind of thing (
> non-invasive, but right amount of functionality)
>
> 5. technical aspects
> - port to Globus 2.0 was relatively easy
>
> Did I omit anything?
>
> __________________________
> Dr. Kate Keahey
> Math & Computer Science Div.
> Argonne National Laboratory
> Argonne, IL 60439, USA
> (630) 252-1673
>
> ===============================================================================
>
> This message was sent to the SciDAC National Fusion Collaboratory (NFC)
> workers list nfc-sc01. Visit the Collaboratory at
> <http://www.fusiongrid.org/>.
>
> To unsubscribe from this list, please send a message to
> majordomo@fusion.gat.com with the following text in the *body* of the
> message: unsubscribe nfc-sc01
>
> David P. Schissel: <schissel@fusion.gat.com> <http://fusion.gat.com/~schissel/>
===============================================================================
This message was sent to the SciDAC National Fusion Collaboratory (NFC)
workers list nfc-sc01. Visit the Collaboratory at
<http://www.fusiongrid.org/>.
To unsubscribe from this list, please send a message to
majordomo@fusion.gat.com with the following text in the *body* of the
message: unsubscribe nfc-sc01
David P. Schissel: <schissel@fusion.gat.com> <http://fusion.gat.com/~schissel/>
This archive was generated by hypermail 2.1.1 : Fri Feb 08 2002 - 13:34:29 PST