My current research focuses on designing flexible wireless interconnects for data centers. In the past, I worked on dynamic spectrum auctions that achieve desirable economic properties while adapting spectrum usage to user demand efficiently.
Flexible Wireless Links for Data Centers
Recent measurements have shown that today's data centers are often limited by the sporadic congestion loss caused by traffic bursts. Compared to the complexity and cost of modifying data center architectures, one more attractive approach is to augment wired links with flexible wireless links in 60 GHz band. Existing 60 GHz wireless proposals, however, are limited by link blockage and radio interference problems. We address these key limitations by proposing 3D beamforming as a novel wireless primitive. It bounces the signals off the ceiling, thus extending link connectivity and reducing link interference. Our first paper appeared in HotNets'11. We further built a experimental testbed to verify our design concept, and our work appeared in SIGCOMM'12.
Trading Platforms for Dynamic Spectrum Distribution
We propose dynamic spectrum auctions to distribute spectrum in a short-term basis driven by the dynamic user demands and willingness to pay. The two key enablers of such systems are distribution efficiency and economic robustness. We achieve these two goals by addressing the following challenges: 1) how to allocate spectrum in real-time while addressing the interference constraints, and 2) how to prevent market manipulations and achieve robustness from economic perspective. We published our works in SDR'08, MobiCom'08, INFOCOM'09, MobiHoc'10, INFOCOM'11.
Please refer to MERCURY project website for more details.