The ADvanced Systems Laboratory (ADSL)
Overview

Our group is interested in developing new technology in operating systems and storage. One of our primary design philosophies has been to build new systems and services with gray-box techniques. In doing so, we take the pragmatic view that systems are no longer built in isolation; rather, one often must design and implement a component of a larger system. Thus, the question becomes: how can one gain information and even control other components within the system, when no explicit interface to do so exists?

We have been exploring this question in the realms of operating systems and storage. For a taste of our operating systems research, check out these papers on an empirical approach to disk scheduling, a cool tool called Dust that lets you figure out what your buffer-cache replacement policy is, a new approach to exposing information to applications from within the OS called the infokernel, and a simple, powerful, and safe extension to TCP known as icTCP.

For a sense of our research in storage systems, you might read these papers on semantically-smart disks, a gracefully degrading disk array known as D-GRAID, a smarter second-level caching scheme known as X-RAY, a tool that can extract detailed information from a RAID array called Shear, and a philosophical discussion about Life or Death.

We've also done some work in the domain of wide-area batch computing, including a batch-aware distributed file system BAD-FS, a study of application workloads for these type of environments, and our earlier work on the NeST storage server.

Another area of interest is the 'soft underbelly' of commodity operating systems: those problems that are so entrenched and endemic that they have evaded prior research. An example of this is the Nooks project to improve reliability by isolating device drivers. These problems exist because of the huge scale at which common operating systems are deployed, the heterogenous platforms upon which they execute, and the wide variety of tasks they are called upon to perform.

For a complete list of papers, please see our publications page.