Improvements and Reconsideration of Distributed Snapshot Protocols

  • Adnan Agbaria
  • Published 2006 in
    2006 25th IEEE Symposium on Reliable Distributed…


Distributed snapshots are an important building block for distributed systems, and, among other applications, are useful for constructing efficient checkpointing protocols. In addition to the imposed overhead of the existing distributed snapshot protocols, those protocols are not trivially applicable (if at all) in many of today's distributed systems, e.g., grid, mobile, and sensors systems. After presenting the shortages and the inapplicability of the most popular existing distributed snapshot protocols, this paper discusses improvement directions for the protocols. In addition, it presents a new and an important improvement for the most popular distributed snapshot protocol, which was presented by Chandy and Lamport in 1985. Although the proposed improvement is simple and easy to implement, it has significant benefits in reducing the software and hardware overheads of distributed snapshots. Then, the paper presents proofs for the safety and progress of the new protocol. Lastly, it presents a performance analysis of the protocol using stochastic models

DOI: 10.1109/SRDS.2006.26

9 Figures and Tables

Cite this paper

@article{Agbaria2006ImprovementsAR, title={Improvements and Reconsideration of Distributed Snapshot Protocols}, author={Adnan Agbaria}, journal={2006 25th IEEE Symposium on Reliable Distributed Systems (SRDS'06)}, year={2006}, pages={155-164} }