Single node failure repair method for distributed storage system based on genetic algorithm and MSR code

Abstract

Abstract In view of the minimum storage regeneration (MSR) single fault node repair plan, design a kind of new single repair fault node topology. Consider link bandwidth available and the node processing power of heterogeneity, the repair delay and repair traffic node of combinatorial optimization. Topology can be converted to Steiner tree model with constraints, and design the corresponding hybrid genetic algorithm to solve the approximate global optimal solution. Finally, a compromise between the two optimization objectives is achieved. Simulation results show that under the same MSR code size, the repair delay of this topology is only 60%-80% of that of the traditional tree topology and 30%-40% of that of the traditional star topology. While the repair traffic for this topology is 9% to 65% higher than for the traditional star topology, the repair traffic is only 40% to 55% higher than for the traditional tree topology.