Geo SCADA Performance Counters

[rlao]

I've been having a play with Geo SCADA performance counters and have successfully gotten them to update the corresponding OPC tags under the PerformanceStats aggregate.

Just wondering what is the recommended way to use these on a multi-server setup (Hot standby w/ Permanent Standby)? So far, I've tried the following:

• Created calculated points to periodically sample the PerformanceStats tags. This pulls performance stats from whatever the main server is at the time. The per-second tags don't seem to be too useful for this, instead I read the the totals instead and subtract consecutive samples to generate per-minute stats.
• Read the PerformanceStats tags directly as mimic expressions. This seems to pull performance stats from whatever server the ViewX client is connected to. So different ViewX clients will see different numbers if they are connected to different servers.

Also, should the Geo SCADA counters be registered on permanent standby servers? Or is this really only useful for main servers?

Cheers

[BevanWeiss · Accepted solution]

I've never actually used the Geo SCADA Expert Performance Counters (although they're quite a recent addition also).

Some of the counters (like Query Performance) would be applicable on Permanent Standby servers, but likely it wouldn't be very exciting, unless you have Performance Servers which have significant load applied to them.  But then likely the existing Windows Performance Monitor stuff (CPU usage, Memory, disk queue etc) would be a more generic metric of 'server load', which is probably more important to performance than just the Geo SCADA Expert query loading (and arguably you could refine the Windows Performance Monitor stats to just be specific Geo SCADA Expert processes to get QP load etc).

I think your first option is probably the best way to go if you want a display (or history) of the Main server loading.

This is likely the best path for being able to look at regressions in Test/Dev activity also.  If a change is about to be imported, but it shows a significant uptick in Queries / Events etc etc then it could be re-assessed as to whether it provides benefits proportionate to it's "cost" (i.e. if it will double your server loading, but it just provides an animated spinning circle.. maybe not worth it) 

If you care about showing individual server loads, then you might need to get tricky and use the OPC driver to retrieve the data from the other servers and log to OPC points.

I can't see much use for displaying just the current server stats.  It would only be if you had a number of servers clients connected to, and you wanted an easy way for them to report the loading on the server they are connected to if they are complaining about bad performance.  But again, it would likely be better to refer to Windows Performance Monitor data then... since it's a better metric of overall server performance.

[rlao]

Thanks, I've had them running for a month now on all our primary/standby servers with no issues and getting some good diagnostic data on main server performance. Can definitely say they are quite handy.