Check if the CSV is owned by the same cluster node that owns the VM. If not, move one of the two over and see if that makes a difference in the performance that you're measuring.
CSV allows you to have the disk (volume) owned by one host (physical server) while a guest (VM) on there is owned by a different host. To allow this to work uses the CSV network link. If this link is slow, your disk access is slow. Moving the guest and CSV onto the same host means it doesn't have to go through the CSV link to be able to write to the disk, and therefor it may be faster.
This is why I asked if you had verified if which host owns the CSV and if that's the same host that owns the guest that you're testing performance on.
Sorry...should have elaborated. The reason I copied the VM to the other host is to rule out a host issue...so it was tried on the one that hosts the CSV as well as the other. So each host has a VM on the CSV that is slow.
In the deployment gride, in Table 3, it lists the following: "Best practice recommendation requires additional adapters on each server in the cluster for CSV related traffic", and in the FAQ there is a link to the Microsoft validation wizard.
I think your problem relates to the "bandwidth" and not the array. In the guide I referrenced, it clearly states the following:
The Hyper-V™ server should have two or more network adapters depending on the use cases for the NICs. One dedicated NIC should be configured for remote management of the physical server and one or more NICs should be configured for virtual network traffic. Additional network adapters will be required when connecting to iSCSI SAN volumes and Failover Clusters including a dedicated NIC for VM Live Migration if using CSVs.
After reviewing this informaion, please post any other question you might have.
If it was bandwidth, I would think I'd see the issue on the non CSV volume as well as the CSV one no? We have 8 Gigabit NICs in each server...2 in each are dedicated to iScsi (stacked dedicated 6224 switches for iScsi traffic), 3 are for the VMs, 1 for cluster and one for live migration. All recommended by Dell for this setup and software parts configured remotely by Dell.
Can you private message me you name, company name, email address, EqualLogic case number and your TAM name (if known). I would like to review the history of the issue first, and if needed, get a support engineer to discuss this with you further.
To private message me, add me to your friends list on this forum, then you will see the option to email me.
Dev Mgr
4 Operator
•
9.3K Posts
0
February 11th, 2011 17:00
Check if the CSV is owned by the same cluster node that owns the VM. If not, move one of the two over and see if that makes a difference in the performance that you're measuring.
brucefan
16 Posts
0
February 14th, 2011 06:00
The cluster is only 2 servers so I copied the CSV VM onto both hosts with just different hostnames...no difference.
Dev Mgr
4 Operator
•
9.3K Posts
0
February 14th, 2011 07:00
CSV allows you to have the disk (volume) owned by one host (physical server) while a guest (VM) on there is owned by a different host. To allow this to work uses the CSV network link. If this link is slow, your disk access is slow. Moving the guest and CSV onto the same host means it doesn't have to go through the CSV link to be able to write to the disk, and therefor it may be faster.
This is why I asked if you had verified if which host owns the CSV and if that's the same host that owns the guest that you're testing performance on.
brucefan
16 Posts
0
February 14th, 2011 08:00
Sorry...should have elaborated. The reason I copied the VM to the other host is to rule out a host issue...so it was tried on the one that hosts the CSV as well as the other. So each host has a VM on the CSV that is slow.
Joe S586
7 Technologist
•
729 Posts
0
February 14th, 2011 09:00
BruceFan,
Hi, I'm Joe with Dell EqualLogic. Not sure if you had a chance to review the following:
http://www.equallogic.com/resourcecenter/assetview.aspx?id=7787 (Deploying Microsoft® Hyper-V™ with Dell EqualLogic™ PS Series Arrays)
And the FAQ: http://www.delltechcenter.com/page/Hyper-V+R2+CSV+FAQ
In the deployment gride, in Table 3, it lists the following: "Best practice recommendation requires additional adapters on each server in the cluster for CSV related traffic", and in the FAQ there is a link to the Microsoft validation wizard.
I think your problem relates to the "bandwidth" and not the array. In the guide I referrenced, it clearly states the following:
The Hyper-V™ server should have two or more network adapters depending on the use cases for the NICs. One dedicated NIC should be configured for remote management of the physical server and one or more NICs should be configured for virtual network traffic. Additional network adapters will be required when connecting to iSCSI SAN volumes and Failover Clusters including a dedicated NIC for VM Live Migration if using CSVs.
After reviewing this informaion, please post any other question you might have.
Regards,
Joe
brucefan
16 Posts
0
February 14th, 2011 10:00
If it was bandwidth, I would think I'd see the issue on the non CSV volume as well as the CSV one no? We have 8 Gigabit NICs in each server...2 in each are dedicated to iScsi (stacked dedicated 6224 switches for iScsi traffic), 3 are for the VMs, 1 for cluster and one for live migration. All recommended by Dell for this setup and software parts configured remotely by Dell.
Joe S586
7 Technologist
•
729 Posts
0
February 15th, 2011 07:00
Can you private message me you name, company name, email address, EqualLogic case number and your TAM name (if known). I would like to review the history of the issue first, and if needed, get a support engineer to discuss this with you further.
To private message me, add me to your friends list on this forum, then you will see the option to email me.
Regards,
Joe