Self-stabilization overhead: A case study on coded atomic storage

Shared memory emulation on distributed message-passing systems can be used as a fault-tolerant and highly available distributed storage solution or as a low-level synchronization primitive. Cadambe et al. proposed the Coded Atomic Storage (CAS) algorithm, which uses erasure coding to achieve data redundancy with much lower communication cost than previous algorithmic solutions. Recently, Dolev et al. introduced a version of CAS where transient faults are included in the fault model, making it self-stabilizing. But self-stabilization comes at a cost, so in this work we examine the overhead of the algorithm by implementing a system we call CASSS (CAS Self-Stabilizing). Our system builds on the self-stabilizing version of CAS, along with several other self-stabilizing building blocks. This provides us with a powerful platform to evaluate the overhead and other aspects of the real-world applicability of the algorithm. In our case-study, we evaluated the system performance by running it on the world-wide distributed platform PlanetLab. Our study shows that CASSS scales very well in terms of the number of servers, the number of concurrent clients, as well as the size of the replicated object. More importantly, it shows (a) to have only a constant overhead compared to the traditional CAS algorithm and (b) the recovery period (after the last occurrence of a transient fault) is no more than the time it takes to perform a few client (read/write) operations. Our results suggest that the self-stabilizing variation of CAS, which is CASSS, does not significantly impact efficiency while dealing with automatic recovery from transient faults.