A Survey on Malleability Solutions for High-Performance Distributed Computing

Maintaining a high rate of productivity, in terms completed jobs per unit time, High-Performance Computing (HPC) facilities is cornerstone the next generation exascale supercomputers. Process malleability presented as straightforward mechanism to address that issue. Nowadays, vast majority HPC are intended for distributed-memory applications based on Message Passing (MP) paradigm. For this reason, many efforts Interface (MPI), de facto standard programming model. Malleability aims rescale executions on-the-fly, other words, reconfigure number and layout processes running applications. involves resources reallocation within system, handling application, redistributing data among those resume execution. This manuscript compiles how different frameworks process malleability, their main features, integration resource management systems, they may be used user codes. paper detailed state-of-the-art devised an entry point researchers who interested malleability.