New and emerging solid-state non-volatile (NV) technologies such as NAND Flash and Phase Change Memory create a new class of high-speed storage devices, both as traditional I/O systems and as extensions of system main memory. To make best use of these technologies in tomorrow's computers we need to be able to model and understand their behavior in such systems, and create algorithms and architectures which leverage their strengths while compensating for their weaknesses.
This work (a) investigates models and algorithms for system behavior of solid-state persistent memory technologies, to achieve performance, reliability, and functionality not possible with today's techniques, (b) examines alternate models for OS-to-storage interfaces to make best use of NV technology characteristics, and (c) researches new operating system algorithms for managing the novel memory model presented by mixing NV and DRAM technologies at the main-memory level within a system.
The coming transition from disk to low-latency solid-state storage as the primary persistent media in computer systems presents an opportunity for transformative changes in the way applications make use of storage; the goal of this research is to achieve such changes, rather than the incremental improvements possible via the disk emulation strategy used with these technologies to date. Research results will include open-source software and hardware; the project also involves development of curricula for undergraduate operating systems education and student involvement at the secondary, undergraduate, and graduate level.
