My research group, The Non-Volatile Systems Laboratory, has two goals: To produce world-class researchers and to be a world leader in understanding how memory technologies will impact future computing systems. We are always looking for new students, post-docs, and visitors to help realize these goals.

Building Systems

The NVSL builds real systems to identify, understand, and solve real problems.  Our systems typically involve hardware, software, and the boundary between them.  Building real systems means our work has impact.

Building Researchers

The NVSL trains students to frame world-changing research questions, devise solutions, and build real systems that solve important problems. You will also learn how to convey those solutions to your peers, other researchers, and the wider world. Our work cannot have impact if no one knows about it.

If you are curious about what, exactly, you’ll learn, here are some essays I’ve written about different aspects of doing research in grad school.

After the NVSL

The most important goal of the NVSL is to train great researchers to prepare them for the career they want. Our alumni have gone on to jobs at leading companies, startups, and places in between.  For instance, Laura Caulfield (Ph.D., 2013) joined Microsoft and makes key decisions about to integrate non-volatile memories into the Microsoft cloud.  Yanqin Jin (Ph.D., 2017) is an early-stage employee at a big-data startup.  You can see where some other students have ended up at the bottom of this page.

Post docs from the NVSL have gone on to be faculty at universities across the country.  Yiying Zhang is now a professor at Purdue, Jack Sampson is at Penn State, and Hung Wei Tseng is at NCSU.

Joining the NVSL

If you like building systems, I would love to have you join the lab.  Prospective Ph.D. and master students should apply the graduate program and mention me as potential advisor.  Potential visitors and post-docs should contact me directly.

NVSL Research Directions

We built the NOn-Volatile memory Accelerated (NOVA) file system specifically to expose the performance non-volatile main memory technologies (like Intel’s 3D-XPoint and NVDIMMs).  NOVA applies many well-known file system techniques but with an NVMM twist.  The resulting file system is faster and easier to use than any other file system available.

Six years before Intel released Optane, we built Onyx, the world’s first publicly demonstrated phase change memory SSD.  Onyx and Moneta (Onyx’s cousin) used an FPGA controller and a reengineered software stack to nearly eliminated OS overheads to deliver amazing performance.  For several years, Moneta and Onyx were the fastest SSDs (GB-for-GB) in the world.

Non-volatile memory is blazingly fast but tricky to use safely, so we built NV-Heaps to understand the challenges and design solutions.  We discovered new kinds of bugs and devised ways to prevent them, so that programmers can build fast, reliable, persistent data structures.  Six years later we are still building on the concepts that NV-Heaps pioneered.

Should you trust your SSD with critical data?  Will your it keep your secrets?  Will it work in space?  To answer these questions we measured the performance, reliability, and security of flash memory.  Our results changed how governments and companies prevent data breaches and changed how researchers around the world think about flash memory.

SSDs are miniature computers sitting alongside your data.  Why shouldn’t they help databases (and other systems) process it?  We developed a series of prototype SSDs that integrate complex data processing functions into the SSDs itself.  We added support for databases, key-value stores, web search, and more.

How do you make use of silicon you can’t turn on?  That is a central question facing processor designers as Moore’s Law seemingly comes to an end.  We developed prototype mobile application processor called GreenDroid that leverage “dark silicon” to dramatically reduce energy consumption in smartphones.

Where Alumni are Now