VMware Cloud Community
mlazzareschi
Contributor
Contributor

Exchange performance issues

Hi everyone,

I'm new to this VM thing so here goes.

We have just recently finished virtualizing our entire data center. We are having a big performance problem with our Exchange sever in particular. Since the migration, we have been experiencing a lot of latencey using Outlook. I ran the Exchange troubleshooting assistant and it is saying that there is a bottleneck in the disk I/O. Any thoughts on what can be done to improve this issue?

All of the servers are on their own LUN, the RAID array is RAID 6.

Thanks,

Mike

Reply
0 Kudos
20 Replies
weinstein5
Immortal
Immortal

HOw many HBAs do you have per host? Have spread the load across all available paths to storage? I assume you have installed VMware tools in your VMs -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
Reply
0 Kudos
mlazzareschi
Contributor
Contributor

What do you mean by HBA's? What should i be looking at?

Reply
0 Kudos
mike_laspina
Champion
Champion

Hi,

Can you post the output of

esxcfg-info -r > esxres.log

Just attach the esxres.log file that it creates

http://blog.laspina.ca/ vExpert 2009
weinstein5
Immortal
Immortal

HBAs are the storage adapters you are using to connect to your SAN - the output of ESXCFG-INFO should have that info -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
Reply
0 Kudos
kukacz
Enthusiast
Enthusiast

What storage are you using? Is it the same as before you migrated to ESX? How many users are working with the Exchange?

--

Lukas Kubin

mlazzareschi
Contributor
Contributor

Is this something that i should be running in an SH session?

Reply
0 Kudos
weinstein5
Immortal
Immortal

yes it is something to run in an SSH session -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful
Reply
0 Kudos
mlazzareschi
Contributor
Contributor

We are using a Storevault S500 with 3.3TB of storage. We migrated our exchange server from a dual 2.GHzProcessors, 2GB RAM, everything installed locally on the physical server. We have about 200 users working on excahnge.

Reply
0 Kudos
mlazzareschi
Contributor
Contributor

Ok I ran the command, how do i get to the log file?

Reply
0 Kudos
kukacz
Enthusiast
Enthusiast

OK. How the storage is connected to the VM? Is it RDM or virtual disk stored on the Storevault?

How many drives, which Raid level and what type of disks (seems to be SATA) are used for the Exchange LUN?

--

Lukas Kubin

Reply
0 Kudos
mike_laspina
Champion
Champion

Most of us use winSCP to retrieve files from an esx host.

http://blog.laspina.ca/ vExpert 2009
Reply
0 Kudos
mlazzareschi
Contributor
Contributor

The Storage is connected via iSCSI on Gigabit switches. They are stored as virtual disks. There are 15 SAS drives in the array with a RAID 6 configuration.

Reply
0 Kudos
mlazzareschi
Contributor
Contributor

downloaded and installed. cannot find the log file. i ran the command as root, where should i be looking?

Reply
0 Kudos
mike_laspina
Champion
Champion

In the directory you were at when you ran the command.

You can rerun it with a full path

esxcfg-info -r > /var/log/esxres.log

http://blog.laspina.ca/ vExpert 2009
Reply
0 Kudos
mlazzareschi
Contributor
Contributor

here's the log.

Reply
0 Kudos
kukacz
Enthusiast
Enthusiast

Hmm, 15SAS drives should be enough for 200 Exchange users. Is the array's performance otherwise acceptable? For example cloning a VM stored on the array could roughly show the performance. You can start the cloning, go to the console, type "esxtop", then press "d" to display the current throughput (the 2 columns on right).

Also, as the storage connection is iSCSI, there might be lots of obstacles on the path. Maybe first try pinging the array's portal IP address from ESX console to verify the basic connectivity and latency for the iSCSI path. You can try the command "vmkping -c 100 IPADDRESS" for it.

You can also ask for help your storage service provider (or reseller), there might some known problem with the array's firmware, the Raid6 or some specific settings. Eg. chunksize.

Did you perform any performance testing on the array before placing the Exchange on it? I can send you an IOmeter test definition to benchmark the perfomance.

--

Lukas Kubin

Reply
0 Kudos
mike_laspina
Champion
Champion

Nothing bad on the CPU/MEM side.

Do you have dedicated switches for your iSCSI SAN?

Could you screen shot the output of esxtop showing network and disk?

hit N and D keys to get those outputs.

http://blog.laspina.ca/ vExpert 2009
Reply
0 Kudos
Anders_Gregerse
Hot Shot
Hot Shot

Did you convert your Exchange from a physical server to a virtual server or did you do a reinstallation?

Assuming that you converted the server, what OS was it running. There have been some issues with conversion of physical server with more than 1 cpu in Windows 2000, resulting in poor performance because the converted server was only using 1 vCPU, but had a multiprocessor HAL (you can verify this in the device manager, look for uniprocesser acpi or multi processer acpi)

How many servers is running on the storage system? Is Exchange the only one with a problem?

Reply
0 Kudos
JRink
Enthusiast
Enthusiast

I am using a iSCSI with a Storevault S500 and 12 SATA disks over RAID-DP (sorta like RAID6). I have an Exchange VM with 300 email users and experience no problems whatsoever. I have 6 other VMs on the same SAN device as well.

What do your performance statistics show on your Storevault device through the management software (web interface). There's a section that shows you disk read/writes and gives you an idea how much disk usage is going on.

JR

Reply
0 Kudos