Beyond a theoretical curiosity, sketching algorithms have emerged as promising telemetry and analytics solutions. However, while sketch-based approaches are attractive due to their high resource efficiency and accuracy guarantees, some fundamental challenges remain as stumbling blocks for adoption. In this talk, I will present some of our recent efforts in building practical sketch-based analytics systems. I will discuss the design and implementation of sketch-based analytics that (1) support efficient NetFlow-like multidimensional queries on programmable hardware (e.g., P4) and (2) efficiently scale to parallel settings on end hosts, such as DPDK and eBPF. The developed network analytics solution is the first of its kind deployed in high-performance network processing libraries (e.g., DPDK and Fd.io). Finally, I will end the talk with new research directions in transitioning sketch-based analytics into "prime time" deployment and fulfilling trustworthy and privacy-preserving requirements.
Alan Zaoxing Liu is an Assistant Professor in Electrical and Computer Engineering at Boston University. His work spans computer systems, networks, and applied algorithms to co-design performant, reliable, and secure data analytics solutions across the computing stack. His recent research focuses on designing scalable and trustworthy approximate computing solutions that cut across hardware and software layers to improve large-scale system efficiency and performance. He is a recipient of the best paper award at FAST'19 and received interdisciplinary recognitions, including ACM STOC "Best-of-Theory" plenary talk and USENIX ATC "Best-of-Rest".