Stats in Veeam Monitor vs vCenter?

Real-time performance monitoring and troubleshooting

Stats in Veeam Monitor vs vCenter?

Veeam Logoby stevo » Mon Sep 21, 2009 4:28 am

Hi,

Hoping someone can tell me why I get seemingly completely different from V/Monitor than from vCenter? The metric I'll use as an example is CPU Ready, but all counters seem to be quite different. If I look at the "Last Day" CPU Ready for a VM in V/Monitor I get an average of 0.13% (35.76 ms) but vCentre returns an average from the same period of 543 ms. The latter seems too high to me, so I'd rather believe V/Monitor and report that everything is running sweetly, so I need to qualify which on is right. Real-time/past hour stats for both vCenter and V/Monitor are almost perfectly in sync though, and return CPU ready of around 0.17% and 42.5 ms.

Obviously I think I'm interpreting something wrong?

Thanks in advance for any help.

Steve.
stevo
Lurker
 
Posts: 2
Liked: never
Joined: Mon Sep 21, 2009 2:48 am
Full Name: Stevo

Re: Stats in Veeam Monitor vs vCenter?

Veeam Logoby Gostev » Mon Sep 21, 2009 10:24 am

Hi Steve, when looking at "Last Day" statistics, 543 ms returned by vCenter is CPU ready time for 5 min interval, while 35ms returned by Monitor is CPU ready time for 20 sec interval (so, just more granular statistics). If you multiple this by 15 (20sec * 15 = 5min) you will get about the same value.
Gostev
Veeam Software
 
Posts: 21396
Liked: 2349 times
Joined: Sun Jan 01, 2006 1:01 am
Location: Baar, Switzerland

Re: Stats in Veeam Monitor vs vCenter?

Veeam Logoby stevo » Tue Sep 22, 2009 1:14 am

Hi Gostev,

Thanks for the reply. So really that should mean that in this case, the 20 second interval averaged out by V/Monitor is more correct? We have parties (apps vs ops) now arguing for which value should be considered more correct, i.e. the apps guys are saying that the 5 min interval represents a larger sample so is therefore more representative (like a election poll). We on the other hand say that there's more 20 secs averages stored over the same period, so therefore it's V/Monitor that's providing the more realistic results. I know you're a little biased, but do you agree with the latter interpretation. I suppose it all depends on how the representative data is stored long term.

To put it another way... in you're example above you say to multiply the 20 sec result by 15 to arrive at a similar result to the 5 min average. I'm sure you were just generalising, but can you confirm that the more correct way of doing that calc would be to, add up 15 individual "chunks" of 20 sec intervals. So therefore you'd really need divide both of them and arrive at a figure which is closer to V/Monitor anyway.

Thanks for your help. As I mentioned, I'm really trying to argue in favour of the V/Monitor results so I really appreciate any input that supports this.

Steve.
stevo
Lurker
 
Posts: 2
Liked: never
Joined: Mon Sep 21, 2009 2:48 am
Full Name: Stevo

Re: Stats in Veeam Monitor vs vCenter?

Veeam Logoby Gostev » Tue Sep 22, 2009 10:20 am

Steve, yes you are correct - I was just generalizing. You should add 15 individual subseuqent 20 sec chunks, this way you will get 5 min CPU ready. From VMware administrator's point of view I don't see how having more data may hurt (having 1 average value for 5 min, versus 15 values for the same time span). 5 minutes is eternity in modern computing (my Win7 boot time 15 sec on SSD drive). But I guess application monitoring may have different requirements... it should not be a problem for us to add this option in the future release.
Gostev
Veeam Software
 
Posts: 21396
Liked: 2349 times
Joined: Sun Jan 01, 2006 1:01 am
Location: Baar, Switzerland


Return to Monitoring



Who is online

Users browsing this forum: No registered users and 2 guests