Design of Efficient Network Coding Schemes for Big Data Storage Networks

Konstantin Khomko

Abstract


I propose an algorithm for a system utilising network coded storage that experiences finite storage I/O bandwidth and a probability of channel deletions. I verify that in comparison to uncoded storage that network coding increases service availability. This is illustrated through two measures of performance; external server blocking probability and average throughput. I illustrate that channel deletions have a larger impact, than internal blocking probability, on the considered measures of performance. I show that algorithm design influences system performance, and propose a heuristically efficient algorithm for data transmission in a system that utilises network coded storage, with finite I/O and channel deletions. This work can be used to illustrate the benefit of network coding in lossy networks, which are analogues of real life systems.

Full Text:

PDF