perfSONAR User Q&A and Other Discussion

[Alan Whinery <>]

OK, thanks.

I can speculate as well as any of you about whether or not the issue is simply about system resources, but I really just wanted to know whether someone who was familiar with the code and characteristics would see something I did not.  The flash-swap thing was just a hack to get the install to complete, but in the end, if it makes pscheduler a little slow, but not iperf3 or owping, then it may not matter.

I will investigate whether pscheduler can be optimized to work satisfactorily a small platform. Of Beaglebone Black rev C., I have 50  (on which bwctl-testpoint worked perfectly).  There are questions to be answered still, when resources are non-plentiful, as to the facility of Apache HTTPD server versus something lighter, as to Postgres versus sqlite, or other reductions or optimizations.

One (other) problem with "educational/experimenter" platforms like Raspberry Pi and Beaglebone is that the OSes come with insecure, bloated IoT/web-based-programming environments and other features, which have to be stripped out in order to make a relatively secure, minimal installation in which to put an application context like perfsonar-testpoint. Yesterday's test was done with an unstripped Beaglebone Debian, meaning that there were about a dozen other things competing for resources that needn't. (At least they stopped shipping the OS with a blank root password).

-Alan

On 4/4/2019 2:44 AM, Mark Feit wrote:

Alan Whinery writes:

 

As regards the unable-to-allocate error previously cited from "Processing triggers for pscheduler-server"  on the testpoint install -- 

I did the obvious thing and add swap space, in the form of a 4GB USB stick. The install completes without any error or warning.

 

Most of the BeagleBone models are pretty short on resources and may simply not be capable of running pScheduler.  The Black has a quarter of the RAM we recommend for a test point, and I’d be willing to bet the swap rate, especially to a slow device, would be high enough to cause problems.  There are some other SBCs in the same price range that are less constrained and would probably run perfSONAR just fine.

 

Tests go further, but do not complete. Still working to make sure I'm not missing something before I start pasting results.

 

If you want to send me a couple of samples off-list, I’ll give them a look, but I’m really inclined to believe this may be a case of insufficient machinery.

 

--Mark