I correct, please read again...

[Ramine]

Hello...

I correct, please read again...

Hope you have read my previous post..

And here is an important thing about a USL benchmark..

Read here:

https://www.vividcortex.com/blog/2015/11/13/capacity-planning-with-usl/

And read about the benchmark:

https://www.percona.com/docs/wiki/benchmark:cisco:scale:start

And here is the server configuration:

https://www.percona.com/docs/wiki/benchmark:hardware:cisco_ucs_c250


As you have noticed this guy have used USL to predict
scalability, but i don't think he is right, because
the performance data comes from also hyper-threading,
because the performance data goes up to 32 threads,
so because there is a mix of hyper-threads and threads,
so this have caused USL to not predict correctly the
peak scalability etc. because the regression
can not extract the right information in this scenario,
so you have for example to benchmark with up to 12 thread
to avoid hyper-threading.



Thank you,
Amine Moulay Ramdane.