Electrical Engineering and Computer Science

Networking, Operating Systems, and Distributed Systems

Networking, Operating Systems, and Distributed Systems photo
Computer networks research at the University of Michigan takes a broad, end-to-end view, ranging from wireless networking and mobile computing to the Internet and datacenter networks. At the user-facing side, we are working on redesigning web pages and mobile apps to improve performance and user experience as well as on understanding how to improve the reliability and robustness of infrastructure services to deliver requests to end users. We also cover software/hardware interactions in wireless and mobile networking, focusing on cognitive radio, adaptive networks, spectrum sensing, MAC & network-layer protocols, and mobile applications. In terms of data centers, we emphasize on application-network symbiosis and work on application-aware networking using coflows as well as on network-aware application design.

Operating systems and distributed systems research is tackling exciting new problems that span the gamut of embedded systems, sensor networks, datacenter applications, and Internet-scale cloud services. Operating systems have become pervasive in our daily lives, controlling not just traditional computers and geo-distributed cloud services, but also consumer electronics devices and cyber-physical systems such as automobiles and power grids. Research projects at the University of Michigan are investigating how to make individual computers, data centers, and Internet-scale services more reliable, more secure, faster, more scalable, and easier to manage by applying a range of techniques such as deterministic replay, speculative/redundant execution, erasure coding, and application-infrastructure symbiosis.

As networks become faster, an emerging platform for computing is rack-scale computers (RaSC). We are building software systems for RaSCs that can provide a single machine abstraction out of an entire rack using RDMA-enabled networking. The challenges in terms of efficiency, performance, and programmability arise from the mismatch between software and hardware layers. We are at the forefront of RaSC research, challenging traditional assumptions of single-machine and datacenter-scale computing; in the process, we are discovering new problems and often addressing them for the very first time in systems research.


Related Links


(see profiles)