Optimizing Cauchy Reed-Solomon Codes for P2P Storage Cloud

To overcome the drawbacks in the P2P storage clouds previously proposed, this paper designs a new P2P storage cloud, in which the erasure coding is performed on the dedicated computing nodes, rather than storage nodes. Through experiments, we find that the coding performance of the existing CRS codes has become the bottleneck of the new P2P storage cloud. Thus, we preliminarily optimize the CRS code, and confirm that the hard drive I/O performance in the computing node has caused the bottleneck in CRS coding. Therefore, we propose two buffer-based I/O minimization CRS codes and conduct a performance evaluation. The experimental results show that, on average, the preliminarily optimized CRS codes improve by 43.13%, and the coding performance of the two buffer-based I/O minimization CRS codes reach 202.282 and 275.297 MB/s, respectively, thereby meeting the performance requirements of erasure coding in the P2P storage cloud.