Allocation and scheduling of real-time tasks with volatile/non-volatile hybrid memory systems

Abstract

Computer systems with hybrid memory systems containing both non-volatile memory (NVM) and DRAM enjoy the advantages of being inexpensive, non-volatile, highly dense of NVM and fast access speed of DRAM. In this paper, we study the problem of deploying real-time task systems on a uniprocessor with hybrid memory system. A task has shorter worst-case execution time if it is allocated on DRAM than on NVM. We proposed several algorithms to allocate realtime tasks on NVM and DRAM such that all tasks can meet their deadlines at runtime under EDF scheduling. Simulation experiments are conducted to evaluated the proposed methods. Experiment results show good performance of the proposed methods in both effectiveness and efficiency.