Intf-HybridMem: Page migration in hybrid memories considering cost efficiency

Abstract

One of the great challenges of today's computer memory architectures has been their energy consumption. In this sense, DRAM technology has come close to its scalability limit due to energy issues. Thus, new technologies have emerged intending to replace DRAM as main memory. In this context, non-volatile memories emerge promising to be an alternative for low energy consumption memory systems. Due to the challenges of these new technologies, hybrid approaches have been a trend. When working with hybrid memories, there is a necessity for decisions in which memory modules each data should be stored. This decision must consider the data access profile and the characteristics of memory technology. Thus, this work presents the Intf-HybridMem architecture, a proposal for page migration in hybrid memories using fuzzy systems to support decision making. The fuzzy approach can model the uncertainties of the data access profile and the characteristics of the memory modules. The tests seek to identify the performance limit of the Intf-HybridMem through the use of an oracle mechanism. Results show that there is potential for exploring migration mechanisms with overhead of 0.18% and 0.39% in some scenarios.