Monday, December 20, 2010

Questions and Answers in 60 seconds (Part 2)

-Does R1 allow more than one VM per LUN (and/or didn`t support live migration)?

You could have many VMs per LUN in R1. But when migrating a VM, you actually moved the storage (LUN), which contained the additional VMs. So that would result in some downtime for every VM on that LUN.
In other words: the disk resource was the unit of failover. By this I mean that all VMs stored on a given LUN had to be moved or failed over together.
R2 introduced a significant change in the way storage volumes can be accessed for VMs, and this feature (CSV) is available only for Hyper-V 2008 R2 VMs. CSV functions as a distributed-access file system for access to VHDs. Other cluster technologies from other vendors have accomplished a similar function by creating proprietary cluster file systems. These cluster file systems provide a locking mechanism shared among all the hosts in the cluster that limits access to the disk to a single node at a time, but all nodes have read/writes access. CSV does not use any proprietary volume format, it uses the standard NTFS that Windows has used for years :-). Also: CSV enables all Hyper-V hosts to have full read/write access to the VHDs of the VMs they are hosting. 
CSV is a option that you could turn on in a Failover Cluster that is built with Hyper-V R2 hosts. 
It`s implemented by creating a directory on the C: volume of each node in the cluster.
Example:
C:\ClusterStorage\Volume1
C:\ClusterStorage\Volume2
In this case, two shared disks in the cluster are assigned to CSV. The first volume has 4 VMs stored on it, and the second volume has 5 VMs stored on it. Only one node of the cluster will own the physical LUN of the shared volume, but each volume can be owned by different nodes of the cluster. CSV provides the ability for each node to have full read/write access to the individual VHDs that are used by different VMs.
-Does CSV do anything for performance?
The most impact would be the disk type of your VHDs. Fixed VHDs provides almost the same performance as pass-through disk, but again: gives you the most flexibility. We recommend not to use Dynamic VHDs. Especially in production environment. 
Another thing is to group CSV volumes with similar disk types (as I stated in my previous post. Group SAS disk for one CSV, and SATA for another etc.)
-Where can I get more detailed info on CSV?
Also: Do not hesitate to post things around at the forums. Many people has a lot of experience with Hyper-V, CSV etc.
-Does the hartbeat NIC need IP configured? Can it be on the same switch as the iscsi network?
The hearbeat NIC should as I stated - in an ideal world, have it`s own network. But you could have it over shared networks as well. (Also ISCSI).

No comments: