Performance and effectiveness trade-off for checkpointing in fault-tolerant distributed systems