A Page Cache Management Scheme in Cloud Computing Environments

Abstract

In recent decades, cloud-based shared storage services create a promising solution for data store. In a shared storage, performance isolation and fairness is critical. To boost input-output performance, operating systems employ a kernel managed caching space called the buffer cache or page cache. In a shared storage, owing to the different access characteristics of tenants' workloads, the page cache would be allocated unevenly, leading to unfair performance distribution mto tenants. Therefore, we propose a framework for performance isolation and fairness in a shared storage system. First, the framework explores how to dynamically allocate page cache space among tenants for performance fairness. Then, the framework utilizes credit-based I/O scheduler for performance fairness in storage devices. The experimental results show that, with the proposed framework, interference in a shared storage system can be eliminated and performance fairness can be ensured.