perfSONAR User Q&A and Other Discussion

[Andrew Lake <>]

Hi George,

To add a little to what Ivan said, this is not possible via regulartesting or the meshconfig currently partially due to some of the concerns Ivan laid-out, or at least that perception. If the remote endpoints have adequate limits set though this is less of a concern. What you propose is a perfectly valid thing to do in some contexts. FWIW, in the upcoming 4.0 this gets a little bit easier. Our meshconfig infrastructure won’t directly support setting-up pure third-party tests, but the pscheduler command, which is essentially the bwctl command replacement thats schedules and runs tests, has an —archive option where you can give it the info of an esmond instance and it will store the result there. You can also tell the test to repeat, so you could relatively easily do what you want with a few commands. Taking this a bit further may be something to look at in the future and integrating with the meshconfig, especially since we will have richer options when it comes to limits with pscheduler.

Thanks,

Andy

On December 8, 2016 at 3:42:47 AM, Garnizov, Ivan (RRZE) () wrote:

Hi George,

 

Generally this is currently impossible. Still it is not even  clear, if that should at all be made possible.

From your explanation I understand, that you would like to initiate tests between 2 remote servers.

-          Have you ever asked yourself, if that would be acceptable for the remote parties?

-          Would you be happy that someone initiates tests from your systems on a regular basis?

 

I would like to propose another more considerate approach.

No one (as of my knowledge) brings a perfSONAR measurement point (regardless of the bundle) without collecting the results from scheduled measurements in a MA. With this in mind my suggestion to you is to seek a collaboration with the remote party to gain access to the MA, where the results are collected.

With organizations that rely on public funding that would mean, they can report and demonstrate the interest in the service they are supporting. Even with single deployments (say university tests), an expressed interest would support the effort of maintenance and possible development of the perfSONAR footprint.

Say there is no possible contact there….would you find this instance reliable?

 

Regards,

Ivan

 

 

 

From: [mailto:] On Behalf Of Uhl, George D. (GSFC-423.0)[SGT INC]
Sent: Mittwoch, 7. Dezember 2016 23:01
To: perfsonar-user
Subject: [perfsonar-user] getting perfsonar bwctl results between two 3rd party nodes into esmond

 

Hi,

 

I know this has been brought up in the past but I didn’t see a resolution.  Is there an *easy* way to to get exclusive 3rd party bwctl/iperf3 tests into esmond?  I’m looking at http://software.es.net/esmond/perfsonar_client_rest.html#publishing-data and  I don’t see a simple way to do this without applying some elbow grease.   Being the lazy guy that I am, I was hoping that someone has a readymade script to do this. :)

 

My motivation is to get throughput tests between two 3rd party pS nodes into my maddash dashboard.  I’m running these via an hourly cronjob and outputting the results into a text file.  I’d rather shoe horn 3rd party test results into esmond but I can’t find a way to integrate 3rd party tests into my pS environment via something I can edit – like a local mesh node regulartesting.conf file.  Perfsonar wants one of the test nodes to be an active participant in mesh or a toolkit node that I have access rights to.

 

Since I have no access rights to either 3rd party pS test node the only available option for me (that I see at the moment) is to run the tests outside the pS context (read CLI), massage the output and post it to esmond somehow. Has anybody been able to do this?

 

Thanks,

George