Fusion-cache: A refactored content-aware host-side SSD cache

Abstract

For the merits of high I/O performance and low energy consumption, SSDs have been widely deployed as the host-side cache devices for backend storage to improve the hosted virtual machines’ I/O performance. But in today’s host-side SSD cache, the cache behavior and flash memory are managed individually, without knowing each other’s internal information. This semantic gap not only makes host-side cache inefficient, but also shortens cache device’s lifetime. Additionally, prior studies point out that host-side cache contains much duplicate data. This will further worsen the situation. In this paper, we try to refactor the cache architecture by integrating the management of cache behavior and flash memory to remove the semantic gap. Moreover, data deduplication is applied to improve the usage efficiency of cache device. Compared with conventional host-side caches, experiments show our proposed cache architecture can generate great improvements on I/O performance and device’s lifetime.