Читать книгу From Traditional Fault Tolerance to Blockchain - Wenbing Zhao - Страница 42

An essential requirement for logging and checkpointing protocols is the availability of stable storage. Stable storage can survive process failures in that upon recovery, the information stored in the stable storage is readily available to the recovering process. As such, all checkpoints and messages logged must be stored in stable storage.

There are various forms of stable storage. To tolerate only process failures, it is sufficient to use local disks as stable storage. To tolerate disk failures, redundant disks (such as RAID-1 or RAID-5 [14]) could be used as stable storage. Replicated file systems, such as the Google File Systems [9], can be used as more robust stable storage.