Some Current Research Projects

Below is a sampling of current research projects we are pursuing in my group. I am always open to exploring new ideas, so if you have something you are passionate about, and which you think I can advise you effectively on, then by all means get in touch!

Compound Processes

with Nilanjana Basu

Modern computer software commonly consists of groups of individual programs working together toward the same goal. This common purpose model is not reflected in the way the software actually executes today. Instead, each program runs in carefully guarded isolation, communicating via secure system calls. This project introduces a new concept, a “compound process” which better reflects the realities of modern software. A compound process may host several “guest” programs in a single trust domain, eliminating expensive and often redundant safeguards between components of a single software stack, to yield substantial performance benefits.

Efficient Multi-Threading with Trust

with Ben Baenen

Trust is a Rust-based message-passing framework which aims to replace locking and shared objects with a particularly efficient form of message passing. With Trust, a shared object is entrusted to a single core, which is responsible for all accesses to this object. This ensures race-freedom on the object, eliminates lock contention, and dramatically constrains opportunities for false sharing, often resulting in dramatically higher performance.

Rackwide Computing

with Noaman Ahmad

Building on Trust and its elimination of shared memory computing, Rackwide Computing is the idea that we can program against an entire rack of computers as if it was a single one. With Rackwide computing, we use Trust delegation over RDMA to create a programming framework that seamlessly scales programs to run across multiple machines. In essence, a single process may span an entire rack, allowing (mostly) normal programs to scale to thousands of cores, without the complexities and inefficiencies of cluster computing.