A Transparent, Incremental, Concurrent Checkpoint Mechanism for Real-time and Interactive Applications

A new transparent, incremental, concurrent checkpoint mechanism for real-time and interactive applications called TIC-CKPT has been proposed in this paper, that al-lows the checkpointed process to continue running without stopping while checkpoints are set to a large extent. Through tracing TLB misses to block the first accesses to the target memory pages while dumping memory address space (the most time-consuming step when setting a checkpoint). At that time, a kernel thread called checkpointer copies the target pages to the designated memory buffer for constructing a consistent state of the checkpointed process before resuming the memory accesses. From the experimental results, in contrast to a traditional concurrent checkpoint system, TIC-CKPT saves more than 2 % of the checkpoint time and decreases the stopped time of the checkpointed pro-cess by around 10%. Moreover, the incremental checkpointing functionality has also been implemented in TIC-CKPT. Compared with full checkpointing, incremental checkpointing can reduce the number of copied memory pages more than 69.7 % while the benchmark is the matrix multiplication and the checkpoint intervals are no more than 20 seconds, thus incremental checkpointing can reduce the checkpoint time consequent-ly. Moreover, for setting incremental checkpoints, compared with CRAK incremental checkpoint approach, TIC-CKPT performs a little better while the benchmarks keep the principle of locality