At VMworld 2011, VMware presented a chart that showed their progress in terms of virtualizing various workloads in their own customer base. The chart (shown below) demonstrated that VMware had made some really good progress with some really hard workloads, and mostly excellent progress with easy workloads (low hanging fruit). The interesting question is what is the best way to proceed from here on out.
Virtualization Progress as of Spring 2011
What the chart does not say, but which is of course obvious to us all is that millions of workloads that are not worthy of being called out on a chart like the one below have been successfully virtualized on VMware vSphere. Virtualizing those workloads lead to dramatic server consolidation, which in turn lead to the delivery of a dramatic hard dollar ROI generated by the associated CAPEX savings.
Virtualizing Business Critical Applications
If many (if not most) enterprises have successfully virtualized their low hanging fruit, then obviously what is left are the business critical and performance critical applications. Much has been and is being written about virtualizing these workloads. The Performance and Capacity Management aspects of this issue are covered in our Resource and Availability Monitoring White Paper and our Application Performance Management White Paper (both of which are going to get updated as we approach VMworld 2012).
One thing that is not being discussed nearly as much as the management aspects of virtualizing these applications is the platform requirements for virtualizing these applications. This is because most people are just making an assumption. That assumption is that if VMware vSphere is still the best performing, most scalable, and most manageable virtualization platform on the market and that this happens to be the platform that you are running in your data center that you will in fact choose vSphere to virtualize your business critical and performance critical applications.
This seems like an entirely valid assumption. Certainly of vSphere was good enough for low hanging fruit, and vSphere remains the technical and market leader in virtualization platforms, then vSphere would certainly be the first choice for virtualizing what comes next (which is again those difficult business and performance critical applications).
But What About the Incremental Licensing Costs?
Let’s make an assumption. Let’s assume you are 50% virtualized, and the 50% you have not virtualized are those business and performance critical applications. As you start to virtualize these you will likely discover the following:
- You are not going to get the same kinds of server consolidation ratios for business critical applications that you got with low hanging fruit
- You may bump into the virtual memory licensing aspect of vSphere as business critical applications tend to consume far more memory than do low hanging fruit applications
Therefore even though you have only half of your estate left to virtualize, you might fact the prospect of needed more than twice as many vSphere licenses to virtualize the remaining 50%.
Going Back and Looking at the Low Hanging Fruit
Can you think of a way to get some free vSphere licenses? Well there is a way. If you are a typical large enterprise, you most likely have the license agreement in place with Microsoft that makes Hyper-V free or nearly so to you. So what if you starting migrating low hanging fruit workloads to Hyper-V as a part of the process of migrating business critical workloads from physical server to vSphere? Due to the likely need for more sockets and maybe even some virtual memory licenses, you will probably ending up needing some more vSphere licenses anyway. But you would not need to more than double them just to address the 50% of your un-virtualized workloads.
Managing Hyper-V and vSphere
If you are a typical VMware shop, you certainly have skilled VMware Administrators on staff. What you may not have is the same skill set in managing Hyper-V. This presents the problem of how to managing more than one virtualization platform without recreating the management tool nightmare that you live with over on the physical side. There are three approaches you can take to solve this problem:
- You can simply live with the fact that VMware is managed by vCenter and Hyper-V is managed by SCVVM. However this is probably a really bad idea as it will lead to two entirely different stacks of products that use vCenter and SCVVM which will lead to the tool and process proliferation that destroyed your agility in the physical world.
- You can buy something that claims to let you manage both from one third party console. There are lots of products that claim this. Microsoft even claims that you can manage vSphere from within SCVVM. But as this paper from VMware points out, there are critical things you cannot do, and that means you end up using both vCenter and SCVVM anyway.
- You can use Hotlink to fully manage Hyper-V from within VMware vCenter and let your administrators keep doing exactly what they have been doing.