Consolidating systems with high pages/sec

[neil_murphy]

Hi,

I've been doing some analysis projects for customers using Capacity Planner and I've noticed something that has me a little bit confused. Every time I analyse a comapnies systems, I get a few servers that are excluded from consolidation because they have paging higher than 200 Pages/sec, which is the default threshold in Capacity Planner. I can get around this by changing the threshold value but I'm afraid I'd be hiding a potential issue.

I've used PowerRecon quite extensively also and have not seen this kind of issue.

I'd welcome any comments.

Neil.

[depping]

PowerRecon has got different thresholds than Cap.Planner. Also different metrics for networking etc, which makes it hard to compare value against eachother. I would start with trying to find out why it's paging this much. I wouldn't exclude them from being virtualized but at least move them to the second phase candidates, so you have the time to test it / do some additional research on the applications running on the box.

Duncan

Blogging:

If you find this information useful, please award points for "correct" or "helpful".

[happyhammer · Accepted solution]

Neil, i have been told on the cap planner course to increase the paging parameter upto 10,000 on system thresholds, Recon does not take the paging parameters into account

[neil_murphy]

Thanks Guys.

I did some more searching and found a lot of other people had the same issue. Here is a list of changes to the default thresholds that I found:

CPU Usage 60% (I use 50% personally)

CPU Queue 4

RAM Usage 90% (I leave this at 70%)

Page file usage 100

Paging 10,000

File Cache 2147

Disk Speed 50

Disk i/o 5000

Network 1000

[newnet]

I'm having the same problem, but i still dont know how to change the thersholds, can anyone help me with this!?!?!?!?

[puzzledtux]

I also noticed the paging issue also in the reports. But I was thinking that these parameters will depend on the applications running and the configuration of the esx server hardware. How can one set those parameters to a standard value in each case?

However, I am more worried about VMware's "Industry standard metrics"! Really no one has insight to this and its really hard time convincing the customer to believe these standard values displayed on the performance report chart tab in the dashboard.

Also, one problem I am facing is while assigning the hardware template to a scenario. I can only assign a single hardware template. Suppose I have two phantom server configurations i.e one is high end box and other is a bit lower end server. I would want to generate a report to consolidate into both the servers. I don't see an option to do so

[FredPeterson]

Thanks Guys.

I did some more searching and found a lot of other people had the same issue. Here is a list of changes to the default thresholds that I found:

CPU Usage 60% (I use 50% personally)

CPU Queue 4

These would be very very dependent upon number of CPU's as well as age of CPU's. You could have a 5 year old dual Xeon 2GHz cruising at 75% total CPU - but that could be consolidated down to 1 vCPU on a Nehalem core and it would probably run 50% of the one core and still perform better.

I made those numbers up of course, but you get the gist.