We have a peculiar problem where our application intermittently slows down. Randomly, response time increases to 5+ min and we are struggling to find a pattern or even a root cause.
Is there a way to monitor how many times this is happening per day? I created a multidimensional analysis chart for last 30 days with Response time > 1min. This still gives me a bar graph, each bar is 6 hrs. I need something that starts 00:00 and ends 23:59. The idea is that we are making many minor performance improvements, and I want to show the number of intermittent slowness instances - per day - is gradually coming down.
The starting point is the 'Transactions & Services' data. I need an end graph that shows what I want, it need not be a multidimensional analysis chart.
Don't know the number of requests involved, but would try a list of Purepaths filtered by the slow response time (eg. 1min). That will give you a list of all slow requests. That's where the number of requests comes in: if they are a lot, you might have difficulties approaching the problem from this perspective.
The last 30 days will change the granularity. I think the useful approach could be to check for the last few days with the maximum 1-day timeframe or even less and then correlate the slowness with the hosts' resources.
Secondly, you can take only a sample of slow PurePaths in different timeframes and check the response time hotspots to analyze the root cause behind the said issue.
Clarification. My question is not about purepaths or correlation. I know how to do that, and have been doing that precisely for the last few weeks but no RCA yet.
My question is purely a data representation problem. I want DT to tell me how many times, per day, response time exceeds 1 minute. Preferably in a graph that I can simply copy paste in a report to upper management.