Improving Hypervisor Based SSD Caching with Logically Partitioned Blocks and Scanning in Cloud Environment

摘要

In the era of big data and cloud computing the virtual machine (VM) environment is important where multiple VMs of different operating system and application can be simultaneously run on the same host. In the VM environment the conventional hard disk drive (HDD) has limitations such as low random access performance and high power consumption. Solid State Drive (SSD) is an emerging storage technology, playing a critical role in revolutionizing the storage system design. Recently, SSD storage caching is widely studied for VM-based systems. The existing works on cache space allocation identify the space demand of each VM based on hit ratio. They are not effective for the VMs of shared SSD cache due to the filte ring effect of higher-level caches. In this paper we propose a novel hypervisor-based SSD caching scheme, employing a new metric to accurately determine the demand on SSD cache space of each VM. Computer simulation confirms that it substantially improves the accuracy of cache space allocation compared to the existing schemes. It also allows to display comparable hit ratio as the existing schemes with less amount of SSD cache for the VMs.