Exchange 2010 specification needs.

BoomAM

BoomAM

BoomAM

Soldato
Joined
5 Jul 2003
Posts
16,206
Location
Atlanta, USA
Hi,
Im trying to work out if our VM thats hosting our Exch2010 Mailbox is underspec'd or not.
Basically the spec is:

OS & Data Drives on LUN1 (15k disk array)
Log Drive on LUN2 (same array as above)
10Gb RAM
2x vCPUs
Now performance wise, we dont have a problem, at worst, perhaps a 5 second delay between sending mail between users, and thats very very very rare. Most of the time, within a second or two.
Disk IO is very low, and the memory never fills either.
However the CPU is constantly at 50% load on both vCPUs, and 75% at peak times.
Is it worth giving it another 2x vCPUs or is the above nout to worry about?

Thanks in advance all. :)
 
I would just give it more CPUs unless you are already SERIOUSLY overcomitting.

If the hosts have resources to spare then definitely use it.

EDIT: I always give our live (and even some of our more heavily hit test) exchange/sql boxes 4 vCPUs
 
You also do not mention what CPU your hosts are using.

I am running 3 x Exchange 2007 servers on VMWare, with 1 vCPU.

MS1 - Is our mailbox/transport server this has 16Gb of RAM and barely uses any CPU
MS2 - Is our basic auth CAS server this is for activesync/iphone connection has 8Gb alloacted to it and barely any CPU
MS3 - Is our FBA CAS server, this is where a majority of our store users connect to approx 1500 and that has 16Gb

This is all hosted on the same HP Lefthand with iSCSI Qlogic HBA cards.

Our VM system is hosted on HP Bl460c with 2 x Xeon 5540 CPUs, 48GB Ram.

Kimbie
 
Nikumba said:
You also do not mention what CPU your hosts are using.
Click to expand...
2x Xeon E5405 (2Ghz)

VMWare ESX 4.1

madman045 said:
You dont mention how many users you have etc..
Click to expand...
200ish

Does this server just have the mailbox role?
Click to expand...
Yes

anything on this PDF help?

http://msbmopod.vo.llnwd.net/d1/ppt/TN-091104-AMihm_RSimpson-1032428203-FINAL.pdf
Click to expand...
Not sure, waiting for it to download now. :)


Im only really checking this due to the high idle CPU usage and the fact that a few users, not all, have 'outlook is trying to get data from exchange' errors...
 
Nikumba said:
How many other VMs do you have on the host?
Click to expand...
At the moment, just two, the VM for the CAS and the VM for the HT (Both 2x vCPU/4Gb).
Although medium term, the HT and CAS CPU will move to a second server and that virtual host with the mailbox on it will have 3-5 less intensive VMs on it (Print, 3rd DC, etc; ).
 
So what do we reckon, increase to 4x vCPUs?

Ive got 15mins of downtime of Exchange scheduled at 11am for a server update to Rollup2, so can change vCPUs then...
 
Sounds odd I would not have expected Exchange to use that much resource, when your environment is tiny compared to what I am running.

I accept 2010 is newer than 2007 so you would like to think the newer ver would work better

Be careful with increasing the vCPU as you will run into a stalling problem.

At the moment you have two quad core chips so gives you 8 vCPU's You are on about allocating 4 to your exchange platform, so incresing the use to 6 so not too bad, but if you assign more than 8 vCPUs in total over all your VMs you can get stalling. If Exchange wants to execute something it will wait for 4 available vCPUs.

We discovered this as we applied real world CPU allocation to VMs and in reality we only use one vCPU per system unless it shows a need to have more, this gets away from this stalling issue we have noticed

Kimbie
 
Last edited:
Nikumba said:
Sounds odd I would not have expected Exchange to use that much resource, when your environment is tiny compared to what I am running.

I accept 2010 is newer than 2007 so you would like to think the newer ver would work better

Be careful with increasing the vCPU as you will run into a stalling problem.

At the moment you have two quad core chips so gives you 8 vCPU's You are on about allocating 4 to your exchange platform, so incresing the use to 6 so not too bad, but if you assign more than 8 vCPUs in total over all your VMs you can get stalling. If Exchange wants to execute something it will wait for 4 available vCPUs.

We discovered this as we applied real world CPU allocation to VMs and in reality we only use one vCPU per system unless it shows a need to have more, this gets away from this stalling issue we have noticed

Kimbie
Click to expand...
Stalling?
Ive never heard of this before. Can you elaborate more or link to an article.

Even using your advice as gospel, with 2x VMs using 2x vCPUs & one moving to 4x vCPUs, thats still 8 vCPUs used on an 8 physical core system?
 
I will try to find the article about it.

It might be worth having a word with VM Support or MS Support. I would certainly post on the VM Community forum, but if you do try to detail your setup as best you can, so model numbers of your hosts, amount of ram etc

Kimbie
 
Nikumba said:
I will try to find the article about it.

It might be worth having a word with VM Support or MS Support. I would certainly post on the VM Community forum, but if you do try to detail your setup as best you can, so model numbers of your hosts, amount of ram etc

Kimbie
Click to expand...

Im gonna up it to 4x vCPUs and see how it goes. Can always lower it at a later date.

Most articles/blogs ive read shows Exch2010 Mailbox servers benefiting from 4x vCPUs anyway.

What details would you like?
 
How odd.

Upped it to 4x vCPUs, and CPU usage is still high, this time across all 4 cores!
Performance wise, it appears to perform about the same.
Memory is slowly creeping up as more users connect and their databases, i assume, start caching, but CPU usage is about 25-40% across the board at the moment.
According to the vSphere client, its definatelly taking advantage of the extra cores, as the CPU cycles its using has gone from 2,500 to 4,500.

3 biggest CPU hoggers according to task manager are:
System Idle - whatevers left over.
Store.exe - about 10-15%.
Ekrn.exe - About 15-20% - ESET Mail Security (AV, Spam, Malware).

Does that seem about right?
 
do you have all the relevant exclusions set up in NOD32?

CPU usage you have there i would say is "normal" for the amount of activity it has
 
madman045 said:
do you have all the relevant exclusions set up in NOD32?

CPU usage you have there i would say is "normal" for the amount of activity it has
Click to expand...
Bizarrely, I checked the exceptions for NOD and they were already in(1day old NOD install), so it must come with them as defaults now.

Is that usage 'good enough' for 4x vCPU's? I'm wondering if the number of vCPU's is overkill?
 
You must log in or register to reply here.
Back
Top Bottom