perfSONAR User Q&A and Other Discussion

[Brian Tierney <>]

I think this would be useful too.

I created an issue in github to track this: https://github.com/perfsonar/pscheduler/issues/456

On Thu, May 25, 2017 at 1:51 AM, Tim Chown <> wrote:

Hi,

> On 24 May 2017, at 13:27, Mark Feit <> wrote:
>
> (One email in this thread wasn’t sent to the mailing list, so I’m re-adding it.)
>
> Brian Tierney writes:
>
> Mark can confirm is this is correct, but I think that as long as one of the 2 endpoints has the full pScheduler installed, you can use the --assist flag to help with this too.
>
> It depends on what you’re trying to accomplish.
>
> To run any two-participant test (throughput, simplestream) between systems A and B, all both must have a functioning pScheduler.  A or B can be tasked remotely by system C that has the pscheduler-core package installed:
>
>             [user@C]$ pscheduler task --assist A throughput --source A --dest B
>
> You can also install perfsonar-tools, but that installs more than is needed to talk to pScheduler.  The --assist switch points the CLI at a pScheduler server that can be used to carry out functions like converting the command-line arguments to the JSON expected by the pScheduler API.  (For reasons I won’t go into here, the CLI can’t do that on its own.)
>
> As I said earlier, we don’t have a lightweight way to do basic non-pScheduler-to-pScheduler tests.  It is technically possible and we might consider something like that for a future release if there’s enough demand for it and we have the development cycles.

We’re now deploying perfSONAR nodes on our backbone as reference systems for university sites to test against, and to participate in community meshes. When we explain what perfSONAR is, we do get the question “so can we just run one-off iperf tests directly against it?”, which is not an unreasonable question.

At present we recommend they do a perfsonar-testpoint installation with the optional packages included. It seems intuitive that we should try to avoid tests clashing unnecessarily. But having as clean a way as possible to allow direct iperf tests might be useful, and might help people understand the value of deploying a full perfSONAR system at their site.

Tim

--

Brian Tierney, http://www.es.net/tierney
Energy Sciences Network (ESnet), Berkeley National Lab
http://fasterdata.es.net