perfSONAR User Q&A and Other Discussion

[Aaron Brown <>]

Hey Brian,

That’s due to some specifics of bwctl. You’re correct that it’s running 2 different tests (sending/receiving), but beyond that, the bwctl tests themselves run longer than the duration specified due to slop between the remote host's, and local host’s respective times. This slop gets applied to both the beginning and end of tests since neither side can really be sure what the remote hosts sense of “now” is. Because we can’t calculate the amount of slop a priori, we have to just put in something. Since this is meant to be more a rule of thumb to prevent you from overscheduling, I conservatively did it at 15 seconds (i.e. 7.5 seconds of slop on either side).

Cheers,

Aaron

On Dec 5, 2014, at 10:14 AM, Brian Candler <> wrote:

If I create a throughput test, duration 10 seconds, interval 1800 seconds; and then add two hosts to this (each host has a single IPv4 address); personar says

" Throughput tests will be running 5% of the time"

However 20/1800*100 = 1.1%

Can someone explain the discrepancy here? I can guess that maybe each test is run separately inbound and outbound, which would make 2.2%, but I can't see where the rest comes from.

Also: if I change the test duration to 20 seconds, leaving everything else unchanged, then it says throughput tests will be running 7% of the time.

Thanks,

Brian Candler.