AppBatch I/O Scheduler: Reducing I/O Interferences for Smartphones

• Main Authors: Shang-Xuan Zou, 鄒尚軒
• Other Authors: 楊佳玲
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/42652242776091694952

碩士 === 國立臺灣大學 === 資訊工程學研究所 === 103 === Smartphones have become the dominating computing devices today. The majority of system researches on smartphones focus on improving network middle-ware or application models. Until recently, storage system has started gaining attention. A couple studies show that storage system performance is critical to end-user experiences. In smartphones, applications are switched to background when users execute another new application. However, to re- duce the application-switching-time, smartphones keep the background ap- plications running without really terminating them. Storage accesses from multiple concurrent bac kground applications can still impact the foreground performance. This paper proposes the first I/O scheduler, AppBatch, designed specifically smartphone’s storage system. AppBatch forms requests from the same application into a batch and prioritizes the foreground batch over others. It improves the user experiences of smartphones while preventing the back- ground applications from starving by scheduling batches in weighted-round- robin fashion. The results show AppBatch improves up to 40.3% foreground performance over conventional schedulers on HTC One X smartphone. And the foreground I/O wait time reduction does not come at the cost of back- ground performance.