perfSONAR User Q&A and Other Discussion

[Mark Feit <>]

Andrew Lake writes:

 

…the OWAMP tests (called ‘latencybg’ tests in pscheduler terminology) are a classification of test that can run at the same time as other tests

 

It is possible to use pScheduler to run OWAMP tests that run for a fixed duration, take a break for awhile and run again. This are classified in a way that will not allow them to run at the same time as a throughput test. These are called just “latency” tests in pscheduler terminology.

 

Some additional background:

 

Unlike BWCTL, where all of this behavior was hard-wired, the core of pScheduler doesn’t understand anything about what the tests do.  It does require that each test plugin supply information about how it should be scheduled.  There are four classes for that:  normal, exclusive, background and background-multi.  Normal tests can be scheduled in parallel with anything but an exclusive test.  Exclusive tests are scheduled so nothing other than background tests are running at the same time.  The two background types pay no attention to what else is happening and run no matter what.

 

Throughput, because it has the potential to distort other results, is scheduled exclusively; inert tests like rtt and trace that are less likely to be distorted by throughput measurements are scheduled in the background.   (You can see the a list of how each test is scheduled by running “pscheduler plugins tests”.)

 

Latency is an odd duck because, as Andy explained, there are two flavors.  One-shot measurements like you’d do for throughput are the “latency” test, which is done with the normal scheduling class.  Streaming measurements are the “latencybg” test, which is background-scheduled so it runs continuously, even when exclusive tests like throughput are running.

 

--Mark