Measuring EvaluationCycle

I wanted a clarification on the values returned by “average of evaluationcycle of client” and the Client Usage Profiler results.

Are the values returned based on real-world time, or by the sum of timeslices the client spent using the processor? What I mean is when I get a result like this:

q: ((it / (1000 * 60) ) as string & “:” & (it mod (1000 * 60 ) / 1000) as string & “.” & last 3 of (it as string)) of average of evaluationcycle of client
A: 15:19.234
T: 0.045 ms

Does that mean that the client, on average, completes the evaluation cycle in 15 Real-World Minutes, or does 15 minutes mean “15 minutes of processor time on the system” (which would take much longer than 15 real-world minutes because the client only gets the processor for, say, 200 ms out of every second)?

Likewise, if the client profiler gives a result like
22) 14.544: Enterprise Security.1502235:Background Evaluation
Does this mean that the client evaluated that Fixlet in 14.544 Real-World Seconds, or that it used 14.544 seconds of Processor Time (probably spread out over a minute of real-world time) to evaluate the fixlet?

1 Like

I was just looking at something similar. How I understood the results were the total evaluation time, not CPU time. There is a caveat somewhere about how the time will be thrown off if the machine goes to sleep. Here are the two pieces of relevance I have been using to compare the times.

To get the time as an integer…

average of evaluationcycle of client / 1000 * second / minute

To get it as a time interval…

average duration of evaluationcycle of client

Have you noticed the Enterprise Security Background Evaluations are taking way to long? That is what made me start researching all of this.

Most of my evaluation woes are with my own Baselines (which usually contain 50-100 fixlets from Enterprise Security), but yes, I have seen that some of the Excel 2010 Fixlets in particular are reporting 40 second evaluation times. Once I get a handle on what the measured values actually show I’m considering opening some PMRs on the content relevance.