Parallel algorithms optimization using Actor Model

Abstract

Introduced methods and instrumentation tools for actor model applied to block recursive algorithms optimization. Created formal model of distribution and coordination of the tasks in computation cluster as asynchronous reactive processes with message-passing represented with an actor model and choreography of actors. Created declarative definitions of algorithms which compiles to the system of actors. Proposed scheme of data placement in a cluster using prioritization of block-recursive operations to reduce idling time, data movement, with increased parallelism in situation of high-speed processors and reduced network bandwidth. Implemented adaptive adjustment of the data placement in a cluster at run time to account for current cluster load. Created autotuning of the actor placement in а cluster which uses statistics of previous runs for optimization. Usage of choreography of actors allows to remove central coordinating element and to avoid hard dependencies between cluster nodes, which provides flexible data placement, improves fault tolerance with no single point of failure and allows to use self-healing.