Design And Implementation Of The High-performance Cloud Storage System Based On Erasure Code

With the development of the Internet, the data storage is becoming more and more attention, how to ensure the security of the data in the spotlight, which requires the presence of the distributed data storage systems with safe and reliable, it spread information or several storage nodes in the network, users under sustained effective and highly reliable way to access information, even if the system appears a certain node failure,power blackouts, system failure, or a hacker attacks, still can effectively provide data storageservices.In the current popular distributed storage system, such as Hadoop HDFS, Openstack Swift and so on, by way of data backup are three backup.The advantage is safe, effective and easy to implement.Shortcoming is also very obvious, is to use a lot of storage space.Now a growing Internet development, big IT companies to produce the amount of data every day is probably a PB, if any data using three backup, no doubt a great test of storage cost is put forward.Aiming at this situation, there are a lot of IT companies began to try to EC algorithm is applied to distributed storage, but EC algorithm needs to consume a large amount of CPU and I/O and become the obstacle of the development of the network resources.In this paper, based on the above situation, puts forward some optimization algorithm based on RS code, and implement the code, using C language and assembly last running results show that the optimized code, at the same time, complete the number of code decoding is dozens of times before the code.And, in this paper, the algorithm to replace the existing HDFS three backup strategy, and the file coding test, detection performance than before is also had the very big enhancement.