Software-Distributed Shared Memory for Heterogeneous Machines: Design and Use Considerations

Abstract: Distributed shared memory (DSM) allows to implement and deploy applications onto distributed architectures using the convenient shared memory programming model in which a set of tasks are able to allocate and access data despite their remote localization. With the development of distributed heterogeneous architectures in both HPC and embedded contexts, there is a renewal of interest for systems such as DSM that ease the programmability of complex hardware. In this report, some design considerations are given to build a complete software-DSM (S-DSM). This S-DSM called SAT (Share Among Things) is developed at CEA (the French Alternative Energies and Atomic Energy Commission) within the framework of European project M2DC (Modular Microserver DataCentre) to tackle the problem of managing shared data over microserver architec-tures. The S-DSM features the automatic decomposition of large data into atomic pieces called chunks, the possibility to deploy multiple coherence protocols to manage different chunks, an hybrid programming model based on event programming and a micro-sleep mechanism to decrease the energy consumption on message reception.