Optimizing Regeneration Time by Node Selection in Group Repair Code

Abstract

Distributed storage systems use network coding techniques like replication, erasure codes, local codes, regeneration codes, hybrid code, double code and group repair code to store data efficiently and provide speedy recovery of data during failures. The performance of these approaches is mainly compared on the basis of storage required and repair bandwidth. Out of these, Group Repair Codes is the one that has optimal repair bandwidth for regeneration of nodes. Traditionally, the cost of regeneration was considered to be dependent on the number of nodes participating in the process and the amount of data being transferred. There was not much discussion on the heterogeneity of the network and the capacity of the links between the nodes. In real-time, the nodes are connected to each other with different link capacities due to which the same amount of data takes different duration in reaching its destination. Selecting the node with higher link capacity helps in reducing the data transfer time. So, considering the heterogeneous nature of the network, this paper reduces the regeneration time for Group Repair Codes (GRC). The node selection algorithms for data regeneration have been proposed for GRC and the results of simulation show significant improvement in the regeneration time. Further, the network coding in heterogeneous systems may be explored for factors like network traffic, intermediate nodes, data routing etc.

Keywords

• group repair codes
• regeneration time
• heterogeneous link capacity
• node selection
• replication
• erasure codes