Analysis of which level of synchrony is attainable in each kind of dynamic distributed system

What is it about?

In a dynamic distributed system, its processes may freely leave or join the system. Dynamic systems may consist of a huge amount of processes. Because of this, it is difficult to place constraints on the time needed to transmit a message in these systems. There are multiple types of dynamic distributed systems. This paper analyses whether each of these types may be considered synchronous, partially synchronous or asynchronous when communication mechanisms are considered.

Featured Image

Why is it important?

Highly scalable distributed applications should rely on replication mechanisms. Those applications should be deployed onto a large infrastructure and should deal with varying and potentially high workloads. The nodes or processes that build those systems define a dynamic set since its membership cardinality should be directly proportional to the system current workload. The data managed by those replicated agents should respect a concrete consistency model in order to guarantee the correctness of that service. Many consistency models require some degree of synchrony in order to be supported. Because of this, it is important to know which level of synchrony can be attained in each kind of dynamic distributed system. Thus, depending on the synchrony requirements of each distributed service, this paper provides a guide for deciding on which types of dynamic distributed system that service may be deployed. Besides, the paper describes several mechanisms for dividing a large asynchronous dynamic distributed system into multiple dynamic subsystems with a higher level of synchrony.