HILTI Feb2016




notice1 276

notice2 276

notice3 276

notice4 276



Research Review: Flash Cache Management Policy for Large-Scale Video-On-Demand Storage Server

Research Review: Flash Cache Management Policy for Large-Scale Video-On-Demand Storage Server
1000 until 1100
Meeting Room 7th Floor
Ms Ola Ahmed Al-wesabi
Associate Professor Dr. Putra Sumari
Large-scale video-on-demand (VoD) services use hard disk drives (HDDs) to store large video files. VoD performance is limited by relatively lower access bandwidth and higher access latency of HDD. These limitations increase dramatically when a VoD server receives a large number of simultaneous requests. On the other hand, flash-based solid-state drives (SSDs) are an attractive alternative because they ensure faster access for an intensive reading system such as VoD. Unfortunately, replacing the entire HDD with SSD is costly because of its high cost per gigabyte. Therefore, hybrid HDD storage and the small capacity of a flash-based SSD can provide high but cost-effective performance to the storage subsystem of a VoD server. The main goal is to improve the performance of the VoD server by maximizing the number of concurrent user requests and minimizing startup latency in a cost-effective way. A flash-based SSD stores only popular videos to provide fast access for read-intensive workloads. This study proposes a flash cache management (FCM) scheme that uses a flash-based SSD as non-volatile cache. FCM controls access of multiple requests for the same video file, thereby reducing multiple buffering for the same data. This scheme is executed in two scenarios. The first scenario replaces RAM with a flash-based SSD in a VoD server. This case considers the write workload because of the buffer cache in a flash-based SSD for unpopular videos that are stored in HDD. To avoid the negative effects of garbage collection and erase-before-write, FCM adopts OP-FCL to manage the storage capacity of the SSD during buffering. The second scenario uses a flash-based SSD as cache in conjunction with RAM. The proposed work is expected to satisfy the performance requirements of a large-scale VoD server in terms of input/output bandwidth and access latency.