08-24-2017 11:51 AM
Hi all,
I'd like to get some more information on the key performance metrics (KPMs) in ISE ---- specifically, what units are they using to report some of the metrics?
Specifically, how is "server load" calculated, and how is latency per request reported? I'm pretty sure it's not UNIX server load since mine ranges from 25 to 50.
Thanks!
Justin
Solved! Go to Solution.
08-28-2017 09:39 AM
The data in top is a snapshot at one particular time. The KPM is usually generated and averaged over each hour for the last 24 hours.
Thanks to Craig in correcting me.
KPM data are calculated differently from the raw data points, as below:
ROUND (AVG (load_avg / cpu_count) * 100, 2) avg_load,
ROUND (MAX (load_avg / cpu_count) * 100, 2) max_load
We should probably call them as per-CPU load average percentiles.
08-25-2017 02:13 AM
The metrics are derived from MnT and logs acquired from auth process. For example...
2017-04-17 22:22:34.295 +05:30 0000000341 70000 NOTICE System-Stats: ISE Utilization, ConfigVersionId=66, SysStatsUtilizationCpu=4.87%, SysStatsUtilizationNetwork=eth0: rcvd = 86848\; sent = 4869 \;rcvd_dropped = 0\; sent_dropped = 0, SysStatsUtilizationMemory=77.77%, SysStatsUtilizationDiskIO=2.17%, SysStatsUtilizationDiskSpace=14% /, SysStatsUtilizationDiskSpace=23% /boot, SysStatsUtilizationDiskSpace=2% /storedconfig, SysStatsUtilizationDiskSpace=1% /tmp, SysStatsUtilizationDiskSpace=17% /opt, SysStatsUtilizationDiskSpace=1% /localdisk, AverageRadiusRequestLatency=0, AverageTacacsRequestLatency=0, DeltaRadiusRequestCount=0, DeltaTacacsRequestCount=0, SysStatsUtilizationLoadAvg=0.20, SysStatsCpuCount=4, SysStatsProcessMemoryMB=7926, ActiveSessionCount=0,
That said, I have had an ongoing discussion with engineering to provide better information to clarify the data represented by KPM so that it is more consumable and actionable.
/Craig
08-28-2017 08:40 AM
Thanks. That is helpful, but I do need to know the measurement units for latency and load.
-Justin
08-28-2017 08:52 AM
Latency is in milliseconds.
System load averages is in processes, as it is the average number of processes (for the past 5 minutes on the time of collection) that are either in a runnable or uninterruptable state.
08-28-2017 08:56 AM
Got it. So the system load average is not really load, but the number of processes running in ISE (such as would be shown in the "show application status ise" ?
Thank you,
Justin
08-28-2017 09:02 AM
This is how Linux/Unix represents "loads". In ISE admin CLI, "tech top" shows the numbers. This is not what shown by "show application status ise".
ise-1/admin# tech top
Invoking tech top. Press Control-C to interrupt.
top - 16:01:24 up 33 days, 14:16, 1 user, load average: 1.21, 0.50, 0.39
...
08-28-2017 09:13 AM
The KPM report in ISE doesn't seem to correlate with 'tech top' either.
ISE KPM report shows average load around 20-28.
Tech Top shows load average around 0.66 to 0.75....
So what is the load really saying in the KPM report in ISE?
Thanks!
Justin
08-28-2017 09:39 AM
The data in top is a snapshot at one particular time. The KPM is usually generated and averaged over each hour for the last 24 hours.
Thanks to Craig in correcting me.
KPM data are calculated differently from the raw data points, as below:
ROUND (AVG (load_avg / cpu_count) * 100, 2) avg_load,
ROUND (MAX (load_avg / cpu_count) * 100, 2) max_load
We should probably call them as per-CPU load average percentiles.
12-17-2020 06:23 AM
Hi Justin,
What commands to see the key performance metrics on ISE? here only showing "tech top" command not all information
12-17-2020 07:00 AM
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide