<img alt="" src="https://secure.bomb5mild.com/193737.png" style="display:none;">

Turbonomic Blog

How to Manage VDI Thin Provisioning Tradeoffs

Posted by Kevin Lamb on Nov 7, 2016 8:32:13 AM
Kevin Lamb
Find me on:

Within our data center more times than not we are striving to accomplish two things simultaneously. On the one hand, we want to make sure the applications that are running in our data center are getting the resources that they need. On the other hand, we want to do so with the least amount of money possible. That is, make sure we get the most out of the infrastructure while at the same time assuring application performance. Really that is one of the huge benefits of virtualizing in the first place. With capabilities such as live migration of workloads and high availability, virtualization helped make sure that workloads got what they needed. As for efficiency, virtualization also helped in that regard with overcommit. More specifically thin provisioning—which I would like to talk about today and even more specifically about managing thin provisioning in say a virtual desktop environment.

With thin provisioning there is an inherent risk involved with overcommitting the amount of storage that is provisioned. However thin provisioning has its obvious advantages with improved storage density. Again, another tradeoff. Turbonomic is designed to control and manage those tradeoffs across your data center, including thin provisioning. What about in a virtual desktop environment? What is your Desired State to manage between that tradeoff?

What’s An Appropriate Amount of Thin Provisioning Risk for VDI?

When deploying tens, hundreds or thousands of desktop instances—whatever the number may be—many factors come into play that impact performance on those machines. Storage and managing thin provisioning play a particularly important role with virtual desktops. If you are thin provisioning with VDI you are usually able to over provision much more than in say an environment with virtual servers. In the case of linked clones, each user will get a non-persistent image and storage utilization will most likely not creep up because you are spinning up and spinning down desktops each time a user logs in. Therefore you can provision at a much higher rate than what will actually be utilized within storage array—George Crump does a great job of explaining VDI Storage Challenges. Even with a full desktop image for each user it is unlikely that each desktop will actually fill up all at once causing the array to be maxed out. With VDI you can potentially handle a greater amount of thin provisioning risk whether that is 2x, 4x, 7x or more; it is up to you.

However, as you increase the thin provisioning ratio, the complexity of inventorying and managing storage consumption becomes increasingly difficult. Additionally, thin provisioning can exist at multiple layers. A virtual machine can be thin provisioned relative to its datastore; and a datastore can be thin provisioned relative to its volume. Such configurations - in this case, thin on thin, create efficiency at the expense of significant visibility impact and management complexity.

Turbonomic's End-to-End Approach

As with every aspect of the data center there are constraints and a desired amount of risk based on those constraints. In the case of thin provisioning, there is a different level of risk for different environments. As discussed above, VDI environments usually use over provisioning at a higher percentage due to its architecture. How does one manage that risk even at these varying levels of over commitment? With Turbonomic it is as simple as inputting your target and letting the platform do the rest.

data center congestion

Above, Turbonomic recognizes that there is congestion in the storage provisioned and has found that moving a specific virtual machine (RE-DSL-4) will solve the problem.

Watch a quick summary of how the City of Garland, Texas was able to repurpose over provisioned hardware for their VDI deployment with Turbonomic here.

Let Software Manage the Tradeoffs

The decision engine is not one size fits all, but, rather, takes the constraint as shown in the picture above and drives the right actions to abide by that over commit percentage constraint. Instead of manually and continually checking and interpreting data on how to keep at the tolerable over provisioned amount, Turbonomic's control platform will always keep it from getting to an unhealthy place without compromising on any other resource such as IOPS or latency. Thus the goal of managing tradeoffs, specifically the tradeoffs thin provisioning brings to a VDI environment.

 

Topics: Virtualization

Subscribe Here!

Recent Posts

Posts by Tag

See all