Research On Pattern Erasure Codes

With the coming of the age of big data, the amount of data, storing in data center, is becoming larger and larger. Therefore, the demand of disks, as the main media using in data system, is also larger and larger. Along with increasing number of disks in system, there are a variety of factors, including the disk error itself, environmental factor, time factor and so on, to threat data, security performance of data is becoming lower and lower. In order to ensure the data security, we usually consider adding a few of redundant disks, and encode all data with erasure codes, so as to improve the reliability of data. Storage performance of erasure codes have attracted academia and industry widely to study in recent years.Nowadays, binary array codes have good property, their operation mode is mainly XOR addition operation, the requirement of the system hardware is very low. Although the study of erasure codes has been prevalent, but most of them are general codes, randomly fight against burst errors. However, there are some shortages in studying special pattern erasure codes, such as fixed size codes, pattern codes.In this paper, first, summary and analysis are given about existing erasure codes, which focus on EVENODD codes, Full-2 codes. Then, several special patterns of disk system are studied, including fixed size erasure codes, specific pattern erasure codes. The fixed size pattern codes, based on EVENODD codes, proposes two new family of codes, which code length increased, respectively, one and two more orders of magnitude than traditional codes, and they are more satisfied with large disk system demand which more and more disks are adding into the system. This paper introduces, in detail, their encode, decode, and the key processes give some examples and relevant evidence, and also give related analysis of the performances, including the density of codes, code length and redundancy. The result also compare with the existing erasure code.