We are currently analyzing our PI tag compression settings to determine where we could strike a more optimal balance with performance and data fidelity. To do so we started analyzing our tags using some software called CompressionInsight by Pattern Discovery (an OSIsoft partner). We have found numerous cases where we should be collecting more data.
I would like to conduct some performance testing to provide more insight into how many records we can collect in a day before we start seeing performance issues when returning data to clients such as ProcessBook. For example, I don’t want to collect so much data that we can’t quickly bring up a 3 month trend of a single tag.
My initial thought has been to write some VB behind ProcessBook that will take a handful of tags of various record counts (High, Medium, Low) and cycle through them 1 by 1 automatically, recording the amount of time it takes to return the data at different time ranges. The end result would be a better sense of what the maximum number of records we could comfortably collect per tag per day and still deliver data to clients quickly. Also, this might help identify bottlenecks in our resources when we analyze windows performance counters (recorded with PI Perfmon) during the tests.
Has anyone created such a performance analyzing tool already they would be willing to share? Does anyone have other thoughts\suggestions on the matter?