Scalable infrastructure for next generation data management systems

点击翻译按钮获取中文摘要

The next generation of data management systems will be powered by large scale multicore systems with non-volatile random access memory (NVRAM). Such systems promise unparalleled performance, but are difficult to program because of their highly non-uniform memory architectures (NUMAs). Classical solutions that ignore NUMA issues perform extremely poorly on such systems, so NUMA-aware solutions are needed. Furthermore, NVRAM has not been integrated into system software, and expert knowledge is needed to utilize it effectively. The primary objective of the proposed research is to provide the core data structures and software necessary to drive next-generation data management systems. More specifically, they plan to develop new solutions to classical computational problems, redesign software to exploit the power of NVRAM, and craft new abstractions to expose NVRAM and other unique architectural features to programmers.