Journalling optimization system and method for distributed computations

摘要

A protocol analysis system is provided with data specifying the defined states of processes participating in a distributed computation. State transitions between states are specified as being enabled by (A) receiving a message, (B) unreliably sending a message, or (C) performing an external action such as reliably sending a message. The specification data also identifies process states known to be final states, and all other states are initially denoted as intermediate states. The protocol analysis system determines if any intermediate states can be re- categorized as final states. Then it determines if any state transitions initially identified as unreliable send operations must be treated as derived external actions, and thus made reliable. Thirdly, for each derived external action, the states of the affected application process must be re-evaluated so as to determine if derived final states need to be converted into intermediate states. The resulting determinations as to which states are final states and which messages must be reliable sent are recorded and used to govern execution of the application process. When executing the application process, state transitions entering and leaving intermediate states are normally recorded on stable storage before the state transition is carried out and reliably sent messages are normally recorded on stable storage before being sent. A number of run- time journal optimization techniques reduce the number of state transitions and messages that need to be stored on stable storage.