Paper Info
Title
Replication Management Framework for HDFS Based on Prediction Technique
Abstract
The number of application based on Apache Hadoop is increasing dramatically due to the robustness and dynamic features of this system. At the heart of Apache Hadoop, the Hadoop File System (HDFS) provides the reliability, scalability and high availability to computation by applying a static replication strategy. However, because of the characteristics of parallel operations on the application layer, the accessing frequency for each data file in HDFS is totally different. Consequently, maintaining the same replicating mechanism for every data file might lead to bad effects on the performance. By rigorously considering the drawbacks of HDFS architecture, this paper proposes an approach to dynamically replicate the data file based on the predictive analysis. With the help of probability theory, the utilization of each data file can be predicted to create an individual replication strategy. Eventually, the data file can subsequently be replicated depending on its own access potential. Hence, this approach simultaneously improves the data locality while keeping the analogous redundancy of data storage in comparison with the default replicating scheme.
Year
DOI
Venue
2015
10.1109/CBD.2015.19
2015 Third International Conference on Advanced Cloud and Big Data
Keywords
Field
DocType
Replication,HDFS,proactive prediction,Bayesian Learning,Gaussian Process
Application layer,File system,Computer science,Robustness (computer science),Redundancy (engineering),Data file,Big data,High availability,Operating system,Scalability
Conference
ISBN
Citations 
PageRank 
978-1-4673-8537-4
0
0.34
References 
Authors
25
5
Name
Order
Citations
PageRank
Dinh-Mao Bui1323.35
Thien Huynh-The2294.08
Sungyoung Lee3132.00
Bin Li431830.27
Jin Wang5264.23