Rethinking Memory System Design (along with Interconnects)

Abstract

The memory system is a fundamental performance and energy bottleneck in almost all computing systems. Recent system design, application, and technology trends that require more capacity, bandwidth, efficiency, and predictability out of the memory system make it an even more important system bottleneck [27, 28]. At the same time, DRAM technology is experiencing difficult circuit and device scaling challenges that make the maintenance and enhancement of its capacity, energy-efficiency, and reliability significantly more costly with conventional techniques (see, for example [7, 8, 11, 12, 15, 17, 18, 22, 23, 32]). In this talk, we examine some promising research and design directions to overcome challenges posed by memory scaling. Specifically, we discuss three key solution directions: 1) enabling new memory architectures, functions, interfaces, and better integration of the memory and the rest of the system, including interconnects (e.g., [1, 2, 19, 20, 34-36]), 2) designing a memory system that intelligently employs multiple memory technologies and coordinates memory and storage management using non-volatile memory technologies (e.g., [16-18, 24, 25, 32, 33, 40-42]), 3) providing predictable performance and QoS to applications sharing the memory system (e.g., [3, 9, 10, 13, 14, 26, 29, 37-39]). As we discuss challenges and solution directions in memory, we will point out research opportunities in interconnects and memory-interconnect co-design (e.g., [2, 4-6, 19, 21, 30, 31]).