perfSONAR User Q&A and Other Discussion

[Michael Johnson <>]

Hi Brian,

Thanks for the feedback. Responses below.

On Wed, Jun 14, 2017 at 07:42:16AM +0000, Brian Candler wrote:
> Looking at the graphs from perfsonar 4.0, I notice some changes from 
> 3.5, and I wonder whether all of these are intentional.
>
> (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 haven't heard of this, what makes you think it's a graphing bug and not a 
function of the test schedule? If you can share an example, we can look into 
it.

> (2) It used to be possible to expand the graph further by dragging in 
> an area under the graph.  If there is still a way to do it, I can't 
> see how.

Scrolling up and down, over the graph area, zooms in and out.

> If I update the "start" value in the URL bar, it gets reset to the 
> time 1 day before the end time.

The new graphs are mostly URL-compatible with the old ones. But yes, 
manipulating the start/end times doesn't always work as you would expect. 
There is an open issue on this.

> (3) If I zoom in to the 1 day level, the summary window remains at 300 
> seconds (like it is at the 3 day level).  Previously at this level, 
> the summary window was 0.

I didn't think we changed that summary window, but I could be mistaken. As 
you've discovered, yes, you can override that in the URL. Be careful because 
you could overtax the MA if you try to load a month's worth of raw data, for 
instance. Any choice we make here is inherently a tradeoff between 
performance and chart detail.

> This means that packet loss shows as n out of 3000 packets rather than 
> n out of 600 packets.
>
> I *can* manually update the URL bar to set summaryWindow=0, but since 
> I can't zoom in any further, it's very hard to make out the 
> minute-by-minute detail.

> Maybe the solution to (2) and (3) is just to introduce a new report 
> range on the drop-down, e.g. 4 hours or 6 hours, with summaryWindow 0.

There are also plans to have a more sophisticated way of picking 
date/timerange. For now, does zooming get you more detail?

Thanks,
Michael

--
Michael Johnson
GlobalNOC Software Engineering
Indiana University

812-856-2771

Attachment: smime.p7s
Description: S/MIME cryptographic signature