Adaptations of Block Layout in Distributed Storage Systems

Reliable distributed storage systems implement data distribution in several ways, but roughly considered two main principles are in common use: (i) independent local storages with shared redundancy and (ii) block striping across several storages together with redundancy. This paper is based on an analysis of these two principles. Pointing out advantages and disadvantages in several access scenarios, we propose adaptation procedures between distribution schemes. Several ways exist to adapt data from the independent local storage layout into block striping that differ in two measures - (i) the total cost and (ii) the granularity of block movements. We found adaptation techniques that are a good tradeoff within these two measures. The adaptations can be used to change the data layout into a beneficial one for distributed access scenarios, whereby locally used data is kept in a resource-saving layout.