perfSONAR User Q&A and Other Discussion

[Andrew Lake <>]

On October 25, 2016 at 6:39:49 PM, Kathy Benninger () wrote:

I would like to update management of the XSEDEnet perfSONARs and want to 
understand the implications for the accumulated test data. 

The pSs were set up with manually configured testing several years ago. I 
would like to move to configuration via a central meshconfig file. How will 
test results show up after that change? Will all the old tests/graphing be 
immediately replaced by the newly collected data? Can pre-mesh and 
post-mesh results somehow merge and have some continuity? 

In terms of graphs and/or MaDDash instances you shouldn’t really notice any differences once you make the transition. Assuming the tests are between the same endpoints and you don't do anything like switch a throughput test from TCP to UDP, they should show-up on the same graphs with your old data.

I have a related question on what will happen to old/new data if I separate 
throughput/latency on 10G/1G. 

If you move one of your tests to a new interface, that will change the source address of your outgoing tests and destination of incoming tests, so it will appear on it own graph. The old data will still be in the archive, it will just look like a different test and have its own set of graphs (since technically it is). 

Historical results have been useful and I'd like to preserve that resource 
if possible. 

The mesh configuration will never delete data out of an archive, so the historical data will still be there regardless. Per above, as long as the source and destination are the same and you don’t change the protocol of throughput tests, they should also appear on the same graphs. Even if one or more of those parameters do change, the old data will still be there, it just might show up on a different graph.

Thanks! 

Kathy Benninger 
Pittsburgh Supercomputing Center