Storage I/O Control: Proportional Allocation of Shared Storage Resources

Datacenters have benefited by virtualization which has reduced footprints of datacenters, decreased management cost by consolidating physical resources, eased datacenter management, and offered live migration of computing entities. Virtualization has greatly increased the need for a shared storage infrastructure. Storage systems are accessed by many virtual machines hosted on multiple hosts with varying demands and business priorities. Effective resource management and prioritization among workloads is the key to successful storage consolidation and reduction over-provisioning of storage resources. Existing QoS mechanisms provided by the storage arrays are not sufficient to differentiate I/O priorities at virtual machine level. Research and development at VMware has come up with a way to auto manage and prioritize IO resources among the many workloads sharing common storage. In this talk, I will give an overview of this new concept called Storage IO Control. I will present results of experimental studies conducted to understand the impact of SIOC on application performance. I will also illustrate the dynamic capabilities of SIOC in managing shared I/O resources across an ESX based virtual cluster with performance numbers. Finally, I will present a case study that will underline the need for a fine grained control mechanism for prioritizing I/O resources in a virtual datacenter with shared storage infrastructure and show how SIOC can fill that need

Learning Objectives

Virtualization
Need for a fine grained control mechanism for prioritizing I/O resources in a virtual datacenter with shared storage infrastructure
One of the possible solutions
Benefits of the solution