Quantcast
Channel: VMware Communities: Message List
Viewing all articles
Browse latest Browse all 206069

Understanding and Calculating Co-Stop (CTSP) over multi pCPUs

$
0
0

Hi All,

 

I have a host with 32 physical CPUs (with hyperthreading disabled to make the math easier). How do I determine host level co-stop percentage?

 

From vCenter:

Using Host Co-Stop Values (ie sum of all cores)

Avg Co-Stop for Host: 4500

Duration of Window: 20000ms

 

To go from Co-Stop summation to % I do the following:

Avg_Co-Stop * 100 / 20000

=4500*100/2000

=22.5%

 

Is my host spending on average 22.5% of time in a co-stop state? (Tuning docs say anything above 3 shows CPU contention on the host.)

 

Looking at pCPU Co-Stop Values (ie core average)

Instead of looking at the summation value for the entire host (which was 4500), I'll now assume each pCPU had an average of 4500/32 = 140.


Avg_Co-Stop * 100 / 20000

= 140 * 100 /2000

= 0.7%


Is my host actually spending only 0.7% of time in a co-stop state?

 

As you can see this has a huge affect on determining root cause of performance on vSMP boxes.


Thank you for any insight!



Viewing all articles
Browse latest Browse all 206069

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>