Love mathematics and logic, especially algebra.
Love Free Software. I’m a very fan of RMS.
DIY drone and flight controller.
Bicycling and playing harmonica.
Ph.D. in Computer Science, 2017 - Present
Tongji University
Visiting Assistant in Research in Computer Science, 2018 - 2020
Yale University
B.Sc., M.Sc. in Computer Science, 2012 - 2017
Tongji University
B.Sc. in Mathematics, 2011
Tongji University
A growing trend is that network service providers and applications are integrated more closely through open APIs for better end-to-end performance. In particular, the provided-aided network information can help applications orchestrate their traffic for a better quality of experience, and is gaining a lot of attention from both the academia and the industry. However, enabling automated network-aware application optimization in today’s Internet is difficult because of the multivendor, multi-domain nature of carrier networks. In this paper, we introduce Sextant, a novel network information exposure system, taking a solid step towards enabling network-aware application optimization in carrier networks. Our work is driven by a real requirement from an international ISP, where timely, accurate distance information is critical to fully utilize its CDN caches deployed at different Point-of-Presence (PoP). We design a novel, flexible requirement model, allowing applications to specify their interests. The queries are efficiently carried out using our aggregation and incremental update algorithms. Evaluations demonstrate that our prototype system operates on top of real router software images, and can efficiently handle dynamics from multiple ASes.
As modern network applications (e.g., large data analytics) become more distributed and can conduct application-layer traffic adaptation, they demand better network visibility to better orchestrate their data flows. As a result, the ability to predict the available bandwidth for a set of flows has become a fundamental requirement of today’s networking systems. While there are previous studies addressing the case of non-reactive flows, the prediction for reactive flows, e.g., flows managed by TCP congestion control algorithms, still remains an open problem. In this paper, we take the first step to solving this problem in a data center network. To address both theoretical and practical challenges, we introduce a novel learning-based prediction system based on the NUM model, with two key techniques named fast factor learning (FFL) and efficient flow sampling. We adopt novel techniques to overcome practical concerns such as scalability, convergence and unknown system parameters. A system, Prophet, is proposed leveraging the emerging technologies of Software Defined Networking (SDN) to realize the model. Evaluations demonstrate that our solution achieves significant accuracy in a wide range of settings.