Phasor Protocols Library Hardware Requirements

GPA Products Grid Solutions Framework
1

What should be the minimum hardware requirement to read at least 10 number of PMUs with 60 -100 frames per second ?

** Using the phasor protocols library.

Reading 10 no. of PMUs with 60 frames (1 frame has 50 measurements) in a windows machine with 6 core CPU and 32 GB RAM. And CPU utilization takes more than 50%. Is it a normal ?

2

That does some a little high – are you using the phasor protocol library yourself, directly? Are you using C#/.NET? Or have you just installed openPDC?

Are you doing something with the data once you’ve read it? If you are visualizing the data, UI costs can eat into the CPU.

Also, what is the GHz speed of your CPU?

Just some thoughts…

Thanks,
Ritchie

3

Thanks Ritchie for the response!!!

I am using the library separately with C# and .NET Framework 4.8. (OpenPDC is not used here, just figuring out with best latency approaches).

** So reader part is a stand alone service itself

Only the PMU reader part eats up this much CPU resource.

(Intel Xeon 4310 - 2.10 GHz)

I am processing the measurements(scaling factor included for analog signals). And later as a key Value string (with all measurements defined), I am streaming the data to a broker. (** before streaming, the Key Value string is also logged to file system)

Thanks
Papu

4

Try building the application in Release mode and running it standalone without the Visual Studio debugger attached.

1 Like
5

It was always in Release Mode after few versions of debugging.
The CPU utilization is based on running the .exe file.
I am doubting the multiplication of scaling factor to each measurement as culprit.
total measurements 50 →
50 measurements X 60 frames X 20 devices (testing based on 10 to 20) = 60000 measurements getting processed per second.

(+ The measurement string getting written to the log files after processed)

6

Try temporarily turning off logging to disk, then check CPU. The try turning off streaming to broker, then check CPU. With this you see cost of each major function and determine what can be optimized.

Note that writing to a disk is often the slowest part of any operation.

Good luck!
Ritchie

7

Thanks a lot @ritchiecarroll for your response.

8

Hi Ritchie,
With ref. to your suggestion on disk logging, I removed the disk writing function for every frame, and the CPU usage reduced a bit.

For reading 50 PMU devices (C37.118 protocol) with 60 frames, what should be the minimum hardware requirement in the given context ?

Thanks,
Papu

9

Depends on the hardware, but given your CPU speed is 2.10 GHz, that’s fairly slow. How many cores do you have? If you have several, you can consider doing some work in parallel, e.g., processing each device on a different thread.

Hope that helps!

Thanks,
Ritchie