Flush primitives for asynchronous distributed systems

Synchronous, asynchronous, and causally ordered communication

“…Examples include: flush channels as a weakening of the FIFO-protocol [2]; channels with bounded buffering capacity allowing a variable degree of synchronization freedom from strictly synchronous to totally asynchronous communication by adjusting the so-called synchronization slack between two communicating processes [-29]; multiway rendezvous as a generalization of synchronous message exchange where an arbitrary number of concurrent processes participate in the execution of a single event at the same time [15,20]; remote procedure call abstractions [5,11,34]; and various broadcast and multicast schemes, ranging from weakly synchronized and causally ordered variants to so-called atomic broadcasts [8,39]. Analyzing and discussing the semantics of those mechanisms, however, is out of scope of this article, which x3 This is a slight optimization compared to the solution by Raynal 14 A study with a similar goal [42] was published while our article et al where an extra vector is needed in every process, was under revision.…”

Efficient Algorithms for Distributed Snapshots and Global Virtual Time Approximation

“…In [1], Ahuja presented a similar solution. In order to extend the applicability of the Chandy-Lamport algorithm (and other algorithms) to non-FIFO systems, the author proposes flush primitives for non-FIFO channels.…”

“…Therefore, when process Q receives message (1), it knows that a marker would have arrived if communication were FIFO. Upon receiving a red message, a white process becomes red and does all the actions (i.e., state recording and marker propagation) it would do upon receiving a marker.…”

Units of Computation in Fault-Tolerant Distributed Systems

“…We describe an algorithm to force a ground state assuming F-channels [1]. F-channels support four primitives for sending a message, one of which will be used here.…”