Distributed storage systems (e.g., SAN, iSCSI) are commonly used in the emerging cloud computing systems to provide virtual machine (VM) storage, for efficient storage utilization and fast VM migration. However, as the size of cloud systems and the number of hosted VMs rapidly grow, the scalability of shared VM storage systems becomes a serious issue. Client-side solid-state-based caching has the potential to improve the performance of cloud VM storage by employing solid-state drives (SSDs) available on the client-side of the storage system to exploit the locality inherent in VM IOs. However, there are several key questions to effectively use SSD caches in clouds. First, because of the limited capacity and high cost of SSDs, it is important to determine the proper size and configuration of the caches. Second, because of the diversity of cloud workloads, it is also critical to properly allocate the limited SSD cache capacity among concurrently hosted VMs. This paper provides answers to these questions by studying hundreds of GBs of and months long block IO traces collected from real-world private (FIU) and public (CloudVPS) cloud systems. The overall analysis shows that cloud computing systems are good target for SSD caching, while write-back caching is important to cache performance. The results from CloudVPS and FIU show an average of 74% and 78% hit ratio and 104% and 68% speedup in IO latency respectively. This analysis has also studied the use of dynamic cache allocation based on working-set size (WSS) analysis. The results show that the WSSes of the traces can be accurately predicted online (with less than 2% relative prediction error for 90 percentile) and the cache allocation can be adjusted dynamically to improve the performance of the competing workloads.

Dulcardo Arteaga received is Bachelor's degree in Computer Science from "Universidad Mayor de San Simon (UMSS)" in Cochabamba, Bolivia 2005. He received is Master degree in Computer Science from Florida International University in 2012. Actually He is PhD. candidate working as research assistant at FIU under supervision of Dr. Ming Zhao. His research focus in in block-level caching and virtualization.