Improved Checkpoint / Restart Using Solid State Disk Drives

Fault tolerance and reliability of distributed systems is often achieved through checkpoint / restart mechanisms. Checkpointing frequency and restart delay crucially depend on data throughput and access performance of the storage medium. In this paper we discuss the opportunity to achieve subsecond checkpointing frequencies and restart delays by substituting magnetic hard disk storage with solid state disks and adapting the access algorithms and storage patterns to the novel constraints imposed by solid