Building systems for AI: A tale of two foundations
|Title:||Building systems for AI: A tale of two foundations|
|Date:||April 15, 2019 (Monday)|
|Time:||10:00 am - 11:15 am|
|Venue:||Room 121, 1/F, Ho Sin-Hang Engineering Building, The Chinese University of Hong Kong, Shatin, N.T.|
|Speaker:|| Dr. Hong XU
Department of Computer Science
City University of Hong Kong
The fast-growing AI and big data workloads already empower much of our everyday life, and is set to define our future lifestyle with jaw-dropping new applications on the horizon. Systems research is critical because the recent success of AI and big data is in large part enabled by datacenter-scale computing infrastructures, which employ an army of machines to harness massive datasets in a continuous fashion.
In this talk, I will present my research that focuses on two system foundations to better support AI and big data. First, we build new data intensive systems that execute the data processing pipelines faster with higher resource utilization. Examples include job schedulers for Spark that provide 60% better makespan, and machine learning systems that compress the embedding vectors by over 100x without performance loss for Tencent’s recommendation models. Second, we build new data center network architectures that deliver more performance and flexibility for data communication. Examples include congestion-aware routing that accelerates flow completion times by 2x at the 99%ile tail. From a broader perspective, these solutions show that significant gains can be achieved for AI and big data systems, by exploiting the unique characteristics of upper-layer workloads and the underlying infrastructure. Fresh opportunities await across the boundaries of systems, networking, and machine learning.
Hong Xu is an assistant professor in Department of Computer Science, City University of Hong Kong. His research area is computer networking and systems, particularly machine learning/big data systems and data center networks. He received the B.Eng. degree from The Chinese University of Hong Kong in 2007, and the M.A.Sc. and Ph.D. degrees from University of Toronto in 2009 and 2013, respectively. He was the recipient of an Early Career Scheme Grant from the Hong Kong Research Grants Council in 2014. He received several best paper awards, including the IEEE ICNP 2015 best paper award. He is a senior member of IEEE and member of ACM.
Enquiries: Ms. Tracy Shum at tel. 3943 8438
For more information, please refer to http://www.cse.cuhk.edu.hk/en/events