An improved schema of coordinated checkpointing protocol for distributed systems based on popular process
Zohra Abdelhafidi, Mohamed Djoudi, M. B. Yagoubi · 2012
In this paper, we propose an improved scheme of non-blocking checkpointing algorithm for distributed systems that minimizes the request number. It is based on piggybacking of dependency vectors not on request messages but on computation messages and replies. Here, a process can initiate checkpointing only if it is a popular process (a process that has dependency information percentage greater or equal to decision threshold). We compare our algorithm called NNB (New Non-Blocking) to CSNB protocol (Cao and Singhal non-blocking protocol) using simulation. To evaluate protocols performance, we choose request number, mutable checkpoints number and first phase duration as performance metrics.