ThinkVirt

As I wrote in this article for TechTarget, VMware's partner-tool, RAWC (Reference Architecture Workload Simulator [where's the C?]) is a great tool for stressing a virtual infrastructure before, during or after a VMware View implementations.

RAWC is apparently the brain child of Fred Schimscheimer (VMware - Senior Technical Marketing Manager). Kudos to Fred!

Here are a few scenarios in which I use the tool:

SCENARIO 1
How much hardware (and what type) should the customer buy to support a VDI endeavor?

When I'm involved with a customer during the planning stage of a VMware View solution, I use the RAWC tool to:

• Simulate a load that is a useful percentage (sometimes 100%) of the projected user base
• During the simulation, measure the impact on the compute and memory components of the VI.

For example, I am looking to see how many, "users" I can fit on a given Dell R900, or Cisco UCS blade.

• Measure the impact on the the storage infrastructure. This includes connectivity out to the storage array(s) (e.g. 10GbE, 4Gbs FC,...).

THIS IS VERY IMPORTANT. Many VDI solutions will fail not because the servers weren't spec'd correctly but because the storage was undersized. VDI is far less about storage capacity and far more about IOPS of the storage infrastructure.

• Test the performance of different RAID types.

Attach two LUNs to a vSphere host, each of a different RAID configuration (e.g. RAID 5 and RAID 10). In a large-scale implementation, you may be using different RAID for the different disks of a View implementation. This is outside the scope of this post.

Run a test of 50 View sessions pegging Datastore_7 (which just so happens to be on a RAID 5 LUN). Measure the performance in time to complete, disk impact, IOPS, Mbps, etc.

Now, run the EXACT same test on Datastore_9 (which just so happens to be on a RAID 10 LUN). Measure the performance then compare and contrast to the above.

This is a great way to show the customer how storage design will greatly impact the performance of a VDI solution.

• I also use the above method to test different storage arrays (e.g. NetApp versus EMC) and different storage solutions (NFS versus iSCSI).

SCENARIO 2
I hate VDI!

This is often due to a solution that wasn't designed correctly the first time. The trick here is determining the bottleneck and presenting a solution in a constructive manner.

• Spin up a workload and take a look in vCenter, network switches and SAN monitors to see what tops out first.

Once you figure out the bottleneck (could also very likely be storage) start scaling back the tests until you determine the optimal load for the current design.

Next step is to very likely undertake a redesign phase.

More RAWC goodness to come; this post was fresh from the field.