Storage QoS For Virtualized Environments
As the initial wave of virtualization projects come to a close, many are finding an odd result. CPU utilization is actually lower than when it started. Now users are looking to pile on more workloads but before they do they need to prioritize storage I/O performance to those workloads; they need a QoS for storage.
As the initial wave of virtualization projects come to a close, many are finding an odd result. CPU utilization is actually lower than when it started. Now users are looking to pile on more workloads but before they do they need to prioritize storage I/O performance to those workloads; they need a QoS for storage.The first wave of servers to be virtualized were in many cases low-risk servers with low demands on the CPU, network, and storage resources. In addition, most virtual infrastructure projects started by purchasing new, powerful hosts. The effect; lower overall CPU utilization than what you started with.
The next step is to virtualize servers that are more resource demanding. Using tools available from CiRBA, Tek-Tools, and others allows you to collect information in near real time about these servers and identify eligible candidates for further consolidation. These tools also will point out potential challenges when performing those migrations. One of the common challenges is storage I/O contention.
Contention for storage I/O is a major limiter to a more significant server consolidation result. As we discussed in our article on Storage QoS, what is needed is a way to prioritize the storage I/O channels to insure that priority can be given to specific workloads within the virtualized environment. This QoS priority must be intelligent enough to "follow" the virtual workload as it migrates via capabilities like VMware's VMotion through the infrastructure.
The challenge isn't just prioritizing storage I/O, the same situation exists for IP network I/O as well. Various companies have implemented a form of QoS on 10-GB IP NICs and take advantage of VMware's NetQueue feature. Even FCoE has room in the standard for some level of I/O prioritization, but at this point none of the makers of Converged Network Adaptors are taking advantage of this.
The critical component is that QoS in a virtual environment must be complete. It has to be understood and adhered to from the individual virtual machine to the physical HBA, all the way through to the switch.
As complete QoS begins to become available from interface card and switch manufacturers, the ability to virtualize a higher number of current physical servers becomes a reality. At that point the ROI of server virtualization will reach its maximum potential.
Track us on Twitter: http://twitter.com/storageswiss.
Subscribe to our RSS feed.
George Crump is founder of Storage Switzerland, an analyst firm focused on the virtualization and storage marketplaces. It provides strategic consulting and analysis to storage users, suppliers, and integrators. An industry veteran of more than 25 years, Crump has held engineering and sales positions at various IT industry manufacturers and integrators. Prior to Storage Switzerland, he was CTO at one of the nation's largest integrators.
About the Author
You May Also Like