Hello,
I recently encountered a performance issue with the AF SDK while working on a project, and I wanted to reach out to the community for insights and suggestions.
I have created a sample code (available at https://github.com/seeq12/AFProbe) to investigate the concurrency behavior of the AF SDK. My observations indicate that when using this code to retrieve samples for 4000 attributes on 64 threads, the task completes in 900s. If I load the samples for the same 4000 attributes on only one thread, the task takes 3000s. This suggests that there might be a concurrency bottleneck, as no more than 3.3 requests appear to be served in parallel.
In an attempt to address this, I adjusted the PIArchss_ThreadCount parameter from 32 to 64, matching the number of cores on our data archive servers (32 cores). Unfortunately, this change did not result in the expected improvement.
I'm reaching out to seek guidance and advice from the community. Are there any other parameters or configurations that could potentially enhance the concurrency performance of the AF SDK? I want to ensure that I'm optimizing the system to its fullest potential.
Additionally, if there are any known limitations or factors outside the scope of configuration that may impact AF SDK concurrency, I would be grateful for any insights on that front as well.
Thank you for your time and assistance.
