An Erasure Code-based Approach to Improve Data Recovery and Update Capability

摘要

The growing demand of cloud storage service has raised a great concern to service providers aiming at high service reliability and usability. In this work, we proposed an erasure code-based approach SEDP to improve the capability of data recovery and dynamic update for cloud storage. The coding scheme adopts the main ideology of Shell Sort to maximize the physical distance of data blocks in each coding group which helps recover damaged data blocks in extreme damage conditions. The concept of data page is introduced to simplify and standardize dynamic update procedure. We then designed a smart monitor to patch the cached data pages to the corresponding data blocks to maintain consistency. The patching work is based on two different granularities, time period and modification threshold. These two key parameters are tuned by the monitor according to the information it collects. The prototype is experimentally evaluated in simulated scenarios, focusing on its performance compared with related works and how the monitor improves the performance as the system runs continuously.