Many organizations have adopted a virtualize-first policy, where all new applications brought online are considered for virtualized deployment prior to getting their own bare metal hardware. But there are typically a number of mission critical applications that are still not being virtualized. The hesitation has little to do with stability or reliability in the virtual environment, something that has long ago been proven. The hesitation comes from the inability of IT to assure applications owners that these mission critical applications will get the performance they demand on virtualized systems.
Certainly there are ways to offer that performance assurance. As we found out in our recent webinar “Guaranteeing Virtualized Application Performance” now available on-demand, users have come up with a number of inspired attempts to solve this problem. Two of these are creating performance-specific LUNs and buying performance-specific storage systems. But both have problems. In this column we will take a look at each of these solutions and offer a potential answer to its application performance problem.
Creating Performance Specific LUNs
A performance-specific LUN is a LUN that has a few or only one virtual machine accessing data from it. Doing this removes the random I/O caused by multiple machines accessing the LUN at same time. The problem with this approach is that it increases storage management complexity because the number of LUNs increases dramatically. Also this only solves some of the performance issues because it isolates I/O at the device level. In our webinar we detail all the issues with performance-specific LUNs.
Buying Performance Specific Storage Systems
The other option is to buy a performance-specific storage system for either the entire virtualized environment or for certain virtual machines. This does solve the problem of separating the I/O path for the performance sensitive applications. But the problem with this approach is that it increases storage management complexity because there are now more devices to manage but it also may substantially increase the cost of the infrastructure.
Is The Solution VM Specific QoS?
An increasing number of vendors are offering quality of service (QoS) capabilities for their storage systems. This may not be enough though for a number of reasons. First, if the QoS is being provided by a single system the dedicated controller still becomes a bottleneck. As we discuss in our webinar “Guaranteeing Virtualized Application Performance” the QoS needs to be delivered through a system that is VM-aware and has an abstracted controller design that eliminates bottlenecks.
The key to virtualizing business critical applications is guaranteeing application performance. The problem is that the traditional methods of trying to guarantee that performance only guarantee more cost and complexity of the infrastructure. IT planners should consider storage systems that can offer this assurance with an abstracted storage controller and virtual machine specific QoS. For more information watch our on-demand webinar “Guarantee Hyper-V App Performance With Hyper-V Software Defined Storage”. All on-demand viewers will receive an exclusive copy of our report, “Software Defined Storage, Defined for Hyper-V”