Cost Analysis of the X-code Double Parity Array

The popular RAID5 disk arrays tolerate a single disk failure by using a parity code to reconstruct the contents of a failed disk on demand, but are susceptible to data loss if a second disk fails. The rebuild process which systematically reconstructs the contents of a failed disk on a spare disk may be unsuccessful due to media failures or a second disk failure. Two disk failure tolerant arrays dealing with both problems can be implemented using Reed-Solomon codes or multiple parity schemes such as EVENODD, RDP, X- code, and RM2. All methods incur the minimum level of redundancy in disk accesses and also capacity overhead (except RM2). An appropriate choice of symbol sizes in EVENODD and RDP results in the same access pattern as RAID6 and little disk load imbalance in degraded mode. In this study we consider the load increase and imbalance of the X-code method, since other methods were investigated in previous studies. We derive a general expression for disk loads and present graphs to quantify the load imbalance.