Space4time: Optimization latency-sensitive content service in cloud

摘要

Nowadays, as cloud service is increasingly a commodity, some Cloud Service Providers (CSPs) make profit by providing content services, and play the more important role of delivering content to users. Providing content services presents new challenges for coordination between storage systems and network infrastructures, specifically for latency-sensitive applications such as voice, video, and terminal services. However, prior research has not applied collaboration techniques of storage and network inline to the request path for latency sensitive applications. In this paper, we propose a latency-sensitive content distribution mechanism, Space4time, for a real world system. We observe that operation is largely affected by the collaboration among end users, storage and networks in cloud. Meanwhile, dynamic request routing within the cloud is strongly coupled with content placement decisions when designing the mechanism. Based on blocking probability, we propose content distribution and request routing strategies that take as input storage and network traffic information. Our strategies enable us to balance storage capacity savings and network traffic for performance, as demonstrated in our YouTube trace-based simulation. Our evaluation shows that Space4time outperforms StaticScheme on average 22.3% for access latency improvement With more sites, Space4time has a better performance due to reduced traffic.