Title:Dynamic Distributed Computing and Communication
Abstract: Distributed computing and communication have been studied for decades for classical problems with bounded-size input. This restriction on the input size was motivated by relatively small scale of distributed systems, in which a periodic system synchronization (i.e., agreement on the system state or a snapshot) could be done in relatively small cost, which in turn partitions the data into small and "independent" chunks in-between consecutive synchronization points. Many recent and on-going systems and applications, however, are of large-scale type, in the sense of data flow and the amount of participants. In such systems, periodic system synchronization is very costly, and thus it becomes important to study distributed problems and solutions for unbounded-size data. A few examples have been already shown how to specify such problems and design efficient dynamic solutions to them without using system synchronization. In my talk, I will briefly present the high-level idea of dynamic distributed computing and communication, and compare it with classical approach based on selected problems.