Re: [gt-user] excessive latency

Mon, 21 Jul 2008 12:07:16 -0700 

I was wrong with my statement about RFT - there is of course a file cleanup
when -s is specified. Please check nonetheless if the below helps improving
the performance.

Martin
If the job is the one from below (globusrun-ws -submit -s -F gavosrv1 -c /bin/true) then no file-cleanup with RFT is involved. I found a similar behavior sometimes on one of the ISSGC08 machines, where simple jobs seemed to take a nap between cleanup and done. Unfortunately i didn't follow up on this and cannot reproduce it right now, but i 
remember that the file-system on the headnode was very slow at that time.
The only thing i can think of is the cache cleanup step, where certain files are removed. 

Arthur:
* Is the filesystem on the headnode that contains the user homes very slow (busy NFS or so)? 
* Do you get a better performance if you run
     globusrun-ws -submit -s -F gavosrv1 -f job.xml
  with the following job description stored in job.xml ?

<job>
  <executable>/bin/true</executable>
  <stdout>/tmp/stdout</stdout>
  <stderr>/tmp/stderr</stderr>
</job>


Martin


Charles Bacon wrote:
Interesting - I remember some discussion like that on this list, I think, but what does that rule achieve?
For the original user: delays like that are not normal. The activity in the Cleanup->Done phase is an RFT job that deletes the files associated with the job. Is your GRAM server configured to use a local RFT server? Is the GridFTP server local to the machine running the container? Do you notice slow results using globus-url-copy from the machine to itself? 


Charles

On Jul 21, 2008, at 4:39 AM, Steve White wrote:


Art,

As I understand it, your application runs a single process in the "fork"
job manager.  So you are referring to the latency in running a single
simple process, rather than to that in submission to a batch system.

I now remember that last September, Thomas Brüsemeister pointed out to
us a work-around for a similar problem, at least regarding file transfers. 
It was to add the following 'iptables' rule:
iptables -A OUTPUT -p tcp --syn --dport 113 -j REJECT --reject-with tcp-reset 

We implemented this on many of our systems at AIP, and observed a big
improvement in some kinds of latencey.  Now I see that on some of them
the setting has been lost (after system upgrades, etc.)

Would this improve things for your application?

Cheers!


On 20.07.08, Arthur Carlson wrote:
In the thread "Globus not for real-time application?", a number of users 
discuss whether it is realistic or not to get latencies below 1 second.
Sounds like paradise. I am seeing latencies of up to a minute!

My workstation, gavosrv1.mpe.mpg.de, not the newest anymore, has GTK
4.0.5 installed. When I use globusrun-ws to go from this machine back to 
itself, ... but just look:
[EMAIL PROTECTED] ~]$ time globusrun-ws -submit -s -F gavosrv1 -c /bin/true 
  Delegating user credentials...Done.
  Submitting job...Done.
  Job ID: uuid:52f0f962-54e1-11dd-a56f-0007e914d571
  Termination time: 07/19/2008 15:51 GMT
  Current job state: Active
  Current job state: CleanUp-Hold
  Current job state: CleanUp
  Current job state: Done
  Destroying job...Done.
  Cleaning up any delegated credentials...Done.

  real    0m24.327s
  user    0m1.242s
  sys     0m0.113s
Note that "user" and "sys" times are reasonable. Almost all of this time 
passes between "CleanUp" and "Done". It can't just be checking
credentials because gsissh is done in a jiffy:

  [EMAIL PROTECTED] ~]$ time gsissh -p 2222 gavosrv1
  /bin/true

  real    0m0.649s
  user    0m0.134s
  sys     0m0.020s

Maybe that is already enough for someone to see where the problem lies.
I can also point out that all (at least many) of the machines in our
grid (AstroGrid-D) seem to be affected, but to varying degrees. Here is
a little matrix of tests:

from gavosrv1.mpe.mpg.de to gavosrv1.mpe.mpg.de: 0m27.235s
from gavosrv1.mpe.mpg.de to titan.ari.uni-heidelberg.de: 0m14.324s
from gavosrv1.mpe.mpg.de to udo-gt03.grid.tu-dortmund.de: 0m8.823s

from titan to gavosrv1.mpe.mpg.de: 0m57.208s
from titan to titan.ari.uni-heidelberg.de: 0m16.875s
from titan to udo-gt03.grid.tu-dortmund.de: 0m27.225s

from udo-gt03 to gavosrv1.mpe.mpg.de: 1m5.221s
from udo-gt03 to titan.ari.uni-heidelberg.de: 0m12.905s
from udo-gt03 to udo-gt03.grid.tu-dortmund.de: 0m6.952s

Please tell me I am doing something really stupid. For production of my
application even a minute of latency is not a big deal, but it's a pain
during development and debugging. Right now I am using gsissh instead of 
globusrun-ws just to work around this.

Thank for the lift,
Art Carlson
AstroGrid-D Project
Max-Planck-Institute für extraterrestrische Physik, Garching, Germany



--
| - - - - - - - - - - - - - - - - - - - - - - - - - | Steve White +49(331)7499-202 | e-Science / AstroGrid-D Zi. 35 Bg. 20 | - - - - - - - - - - - - - - - - - - - - - - - - - 
| Astrophysikalisches Institut Potsdam (AIP)
| An der Sternwarte 16, D-14482 Potsdam
|
| Vorstand: Prof. Dr. Matthias Steinmetz, Peter A. Stolz
|
| Stiftung privaten Rechts, Stiftungsverzeichnis Brandenburg: III/7-71-026 | - - - - - - - - - - - - - - - - - - - - - - - - -

Reply via email to