perfSONAR User Q&A and Other Discussion

[Andrew Lake <>]

On June 15, 2017 at 10:01:25 AM, Michael Johnson () wrote:

On Thu, Jun 15, 2017 at 09:13:00AM +0000, Brian Candler wrote: 
>>>(1) If I set a throughput test to run at 1 hour intervals, it 
>>>seems "jerky". Sometimes the interval between tests is 2 hours; 
>>>sometimes it is only 20 minutes. 

I mentioned this privately but wanted to highlight here as well, that the graphs are displaying the data points that were collected, which does indeed happen at somewhat irregular intervals. 

By default, we randomize the intervals somewhere between 1x the interval and 2x the interval (so they run every 1-2 hours in your case). Tests are recreated every 24 hours to prevent orphaned tests if a meshconfig-agent goes away, so my guess is the 20 minute spacing is when one ends and the other begins We have done interval randomization for a number of years for a few reasons, but the biggest being so tests are less likely to stack on top of each other. Without the randomization, your command-line tests are much more likely to have to wait long time to run if you run them at the beginning of your interval vs the end.