The pending arrival of Phase Change Memory: The implications on the memory-storage hierarchy and on future systems development

摘要

Phase Change Memory (PCM) technology has been explored for decades, yet today it is looming on the verge of market impact. Commercially available today for very specific applications, the full scope of value remains to be realized, or even defined. PCM offers unique capabilities and features not seen in any other commercial memory or storage technology, yet its immaturity in a well established marketplace makes adoption in the server space challenging. Full utilization will depend not only upon advancements in PCM density and improvements in performance, but also more fundamental changes in the age old DRAM to disk hierarchy and adoption of a new paradigm in how system architecture is approached - technology/system/application co-development. In the age of systems optimized for specific workloads, truly exploiting the value of PCM will depend upon optimizations being made across the entire system stack, from the technology itself, to the system architecture, and in the software stack.