Enhancement of cache utilization by isolation and thin provisioning of hypervisor cache

Abstract

Storage resources are being consolidated and have led to the increased demand in sharing storage, depending on the type of application or class of customers. In such a scenario, cache consolidation becomes increasingly necessary. An analysis of cache utilization at various levels in a Storage Area Network such as the server cache, the virtual machine cache, and the array controller cache was done, drawing conclusions from the response time and execution time. A proposal is made for dynamic cache allocation algorithm in a virtualization environment, which takes into account the usage trends and varies the allocation depending on the type of the workload. The dynamic cache allocation algorithm helps in adapting the cache space available for a particular virtual machine based on the number of requests and the cache miss affected, with respect to a threshold value which is individual to each VM. The caching mechanism proposed in this work is an isolated, unshared, dynamic cache, where caching is done on the hypervisor instead of the virtual machines. Also, we prove by experimental results that a static allocation of the cache is not suitable for a varying workload in a virtualized setup.