vSphere 4 Exchange Reference architecture is up!

The Solution Validation machine has been churning ever since we got RC1, and then hit full stride when the RTM bits arrived.

So….

Mission Critical Apps on vSphere reference architectures are now starting to roll off the presses pretty fast (next up… Sharepoint, blended Exchange/SQL Server/Sharepoint at very large scale, Oracle 11g).   I can see them all coming (I keep getting asked for review cycles!!!)

Highlight in here was that the 1000 user building block performed better in vSphere, and the FT feature was used on the Mailbox Server roles – it worked flawlessly under load.

There are detailed docs that go together with these, for example, here were the detailed FT findings….

Note we tested up to 4 building block units on the single ESX host – didn’t break a sweat.   We observed zero outage in the Loadgen test.   Direct comment from the wizard (Derrick Baxter) who was doing the tests:  “Typical clustering solutions result in a period of downtime, no matter how small of downtime none until VMware vSphere 4.0 were zero, VMware Fault Tolerance has the ability to enable clustered Exchange of having zero downtime.”

http://www.emc.com/collateral/hardware/technical-documentation/h6237-virtual-infrastructure-clariion-cx4-120-vsphere-reference-architecture.pdf

Keep posted, I’ll link to them as quickly as I can.

Comments

[Justin]

After reading through the document, I take it the data disks for the Exchange VMs were VMDKs?

[Brian]

Question on the interplay (if any) between FT and HA: can the physical vSphere servers be enabled for both, or is it one or the other?

Scenario:
A client has a mix of VMs on two vSphere physical servers. We want to use FT to protect some of them, and HA to protect the others.

[Derrick Baxter]

The datastores were using ESX iSCSI connecting to the CX array.
The Exchange database, log files, and mount points were all using guest Windows iSCSI initiators with Powerpath for MPIO.

In this scenario we had all 4 serves in one vsphere cluster. You can only select the servers which have the proper processor revisions for FT.
Guests using FT will only be on FT servers.
Guests homed on the other VMHA cluster nodes can move throughout all 4 of the clusters. They can still be protected with VMHA/DRS/Vmotion.

You are describing in your scenario exactly what I did.
4 NODES in a vmha Cluster vi4
2 HA nodes capable of FT feature ( new processors ) vi4
4 Exchange MB servers with FT feature
2 HA nodes (vi4)
1 hub/cas
1 ad ( dc/gc )

The hub/cas and dc can move throughout the 4 nodes.
the 4 exchange servers are in FT and are fault tolerant between the two nodes and would not be moving around the cluster.

hth
Derrick Baxter (EMC)

[Todd Muirhead]

The link in the post takes me to a paper that has great details on the configuration. Where is the paper that has the testing results and the graph that you included in your blog post? (I want to see more pretty graphs and numbers - Please)

[Derrick Baxter]

I'll get the URL. Chad may be able to get the vtr link also.

[Chad Sakac]

Todd - thanks - I've posted a link to the detailed Applied Tech guide here:

http://virtualgeek.typepad.com/virtual_geek/2009/05/more-on-exchange-on-vsphere-including-ft.html

[Brian Mullins]

Congratulations on the launch of vSphere 4.0. In the post, you mention that “Typical clustering solutions result in a period of downtime…none until VMware vSphere 4.0 were zero.” I just wanted to let you know that Marathon’s everRun solution, which has been in deployment for 12+ months, provides zero downtime. In fact, Dan Kusnetzky from ZDnet just published an article last week on his blog about The Sullivan Group, who said that they have had zero downtime using Marathon’s everRun, even during a major failure.
http://blogs.zdnet.com/virtualization/?p=933

Argo Global Capital has also spoken about how they have had zero downtime using Marathon. Just wanted to clarify this. Thanks

[Chad Sakac]

@Brian - you're right - I neglected to include other 3rd party tools like Marathon in my statement. I was thinking comparing to Microsoft native options (WSFC, SCC, SCR)

[Term Papers]

Blogs are so interactive where we get lots o informative on any topics nice job keep it up !!

[Choosing SEO Packages]

The link in the post takes me to a paper that has great details on the configuration. Where is the paper that has the testing results and the graph that you included in your blog post? (I want to see more pretty graphs and numbers - Please)

[Attila Bognár]

Why aren't there two switches for redundancy?