During the virtualization roundtable broadcast VMware announced to make a special community group available on VMTN dedicated to messaging (Exchange/domino and other enterprise messaging), which probably will end up in the general section.
Tonight’s roundtable theme focused on enterprise messaging. The major ingredient tonight was Exchange and Storage. Also clustering (exchange) was briefly touched. It was an interesting session and the hour was over before I noticed, I think a three hour session easily could be filled with this topic.
A couple of highlights:
- no performance difference between rdm’s or vmdk’s
- restoring a vm containing Exchange in a SMB scenario, you easily can use VCB or a backup-client from within the VM. VCB now makes use of vss (volume shadow copy services) through the VMware tools that are installed in the VM, which is no default enabled according to VMware. So VCB is now vss integrated. VCB talks to the exchange operator interface, and with vss you can safely backup the entire VM without having to worry about being left with a non-working vm when you restore the VCB backup of a VM on another host after failure.
- in order to meet support criteria, the shared storage in a MSCS must be RDM
- in case a VM booting from SAN this also should be a RDM
- array level backup also needs RDM for support reasons
- In case of high storage loads you should use dedicated lun’s
- Most issues with Exchange either virtual or physical tend to end up in storage configuration issues
- Start a design of a virtual system like it was physical. Start to identify the business requirements and the SLA-requirements, these are the most important factors that will influence the decision about the amount of mailboxes per vm, the use of clustering to mitigate risks, and therefore the amount of vm’s you will need to service your customer needs
- Best practices can be found in the best practices design guide from VMware
- As we speak VMware is busy creating some sort of a sizing tool for a virtual Exchange environment based on combined best practices from VMware and Microsoft.
- some best practices looking at the components and combining it with other parts of the infrastructure are aimed at affinity rules for cluster nodes to lower the risk of loosing an ESX host.
- Clustering was briefly mentioned, especially CCR on Windows 2008 was mentioned, CCR is build on MSCS so in that perspective it likely will be supported, but in the current version of ESX windows 2008 clustering is not supported. For Clustering there is a need for a special driver supporting shared storage and this driver is not included in the current version (ESX3.5) of ESX. CCR does not need shared storage so this cannot be an issue for not supporting it. The issue now is that it has to be extensible tested by QA before it can be supported. In the new release of VI windows 2008 clustering is supported. Win2003 with CCR is supported.
- Looking at how to provision your vm’s with resources, take the Microsoft recommendation take half of that and see how it works out.
But you can also listen to this (see sidebar)…