Optimal task allocation on non-volatile memory based hybrid main memory

Abstract

This paper targets task allocation problem on hybrid main memory composed of non-volatile memory (NVM) and DRAM. Compared to the conventional memory technology DRAM, the emerging NVM has excellent energy performance due to the ultra low leakage power. However, most types of NVMs come with the disadvantages of much shorter write endurance and longer write latency as opposed to DRAM. This paper explores task allocation problems on hybrid memory which consists of energy-efficient NVM and write-endurable DRAM. The objectives of the task allocation include minimizing the energy consumption, extending the lifetime and minimizing the size. The contributions of this work are twofold. First, we design Integer Linear Programming (ILP) formulations that can solve different objectives optimally. Then, we propose three effective polynomial time heuristic algorithms. All the ILP formulations and the proposed heuristics are executed to optimize multiple objectives offline. Experiments show that compared to the optimal solutions generated by the ILP formulations, the proposed heuristics can produce near-optimal results.