Strategy of replica selection based on grey Markov chain prediction model in HDFS
Yanyan Ding · Journal of Computer Applications · 2011
In Hadoop Distributed File System(HDFS) where the network bandwidth and performance of nodes are limited and changed dynamically,the current strategy of replica selection can not adopt the most suitable replica according to the changes of the environment.Aiming at this problem,a new strategy of replica selection based on grey Markov chain prediction model which takes network band width,performance of I/O and storage space of nodes into the comprehensive consideration was proposed to seek a balance between system usability and load balancing.Simulation experimental results prove the validity and practicability of this new strategy.