Roman Dementiev comments

Results 111 comments of


                                            Roman Dementiev

Grafana Shows Wrong Core_Frequency Value, But Data is Right

thanks for the report. This looks like a bug to me..

Grafana Shows Wrong Core_Frequency Value, But Data is Right

this seems to be specific to the way how this pcm data is exposed in prometheus format. The PCM start.sh grafana script (telegraf/inflixdb/grafana stack) shows these metrics correctly.

Grafana Shows Wrong Core_Frequency Value, But Data is Right

I think I have root-caused the issue.

Help needed to monitor / understand socket interconnect / UPI sync traffic with Intel pcm

Hi, could you please compute how many times per second is the shared variable incremented in your tests? This would give a ball-park number of the expected difference in the...

Help needed to monitor / understand socket interconnect / UPI sync traffic with Intel pcm

>Is it fair to say that for each atomic increment the interconnect would have to be used to sync the shared variable to the peer socket memory? no, for an...

Help needed to monitor / understand socket interconnect / UPI sync traffic with Intel pcm

you might want to read this white paper: https://halobates.de/xeon-lock-scaling-analysis-paper.pdf

Help needed to monitor / understand socket interconnect / UPI sync traffic with Intel pcm

I believe reading this discussion will be helpful for you: https://github.com/intel/pcm/issues/64 Summary: it is possible to get a high UPI utilization using available stress tests but getting 100% is not.

Once started pcm fails to monitor UPI traffic after some time

yes, you are correct. We do not support running multiple instances of pcm at the same time.

probes should use named ports rather than port numbers

@ppalucki FYI

IIO is not returning any value

I think the main issue is that ``opCode-*.txt`` files are not available. Did you run "make install" inside the container to let pcm build process install opCode-*.txt files? Not having...