New NSF grant aims to make non-volatile memory technologies more usable
Their team will work closely with researchers at Intel and Microsoft.
NVM devices retain stored information even after the power supply is turned off. Emerging NVM technologies include magnetic random-access memory (MRAM), spin-transfer torque random-access memory (STT-RAM), ferroelectric RAM, phase-change memory (PCM), and resistive RAM.
"Compared to conventional main memory technologies, non-volatile memory (NVM) is emerging as a revolutionary technology that promises close performance and larger storage capacity, while offering data durability," said Huang, an assistant professor of electrical and computer engineering. "We have already seen real NVM devices on the market. However, how to use them in computing systems in a simple and efficient way is still a big research challenge. In this project, we will develop systems and architecture techniques to improve their programmability, performance, and scalability."
Torrellas, the Saburo Muroga Professor in Computer Science, and Huang, his co-PI, will build an easy-to-use and generic programming framework for NVMs that requires minimal programmer involvement.
The team will redesign hardware primitives in the processor and memory hierarchy to minimize the overhead of both memory persistency operations and other operations to enhance system usability. They will also develop distributed data persistency models and apply them to a distributed computing environment with NVM.
The four-year research project, “Cross-Cutting Effort to Make Non-Volatile Memories Truly Usable,” will begin on October 1.