加拿大多伦多大学的李葆春教授近期来华访问，现已安排在5月11日上午9：15分在东5楼210学术报告厅面向计算机学院师生做一个有关对等计算与GPU领域的学术演讲。欢迎广大师生踊跃参加！
李葆春老师简介：
李葆春教授是加拿大多伦多大学（University of Toronto）电机和计算机工程系讲座教授（Endowed Chair Professor）、计算机工程专业主任，IEEE及IEEE Computer Society高级会员、ACM会员；现任Elsevier Computer Networks Journal and ACM/Springer Multimedia Systems Journal编委，曾任IEEE Transactions on Multimedia and IEEE Transactions on Vehicular Technology编委以及ACM/Kluwer Wireless Networks Special Issue on Quality of Service in Heterogeneous Wired/Wireless Networks客座编委，担任包括ACM 多媒体国际学术会议（ACM Multimedia）、 IEEE IWQoS在内的十余个国际学术会议的程序委员会主席、副主席以及联合主席等职务，八十余次担任包括ACM Nossdav、MobiHoc、IEEE INFOCOM、ICDCS、ICNP、ICPP、PerCom等在内的重要国际学术会议的程序委员会委员。李葆春教授主要从事对等计算、网络编码、无线通讯等领域研究工作。在大规模对等媒体服务拓扑构建、节点选择、ISP（网络运营商）资源调度机制上提出开创性的理论与方法，针对网络编码、博弈论对大规模分布式网络模型及优化分析做出具有原创性的贡献。在包括ACM/IEEE Transactions等顶级期刊上发表30余篇高档次学术论文，在网络领域的旗帜会议IEEE INFOCOM上共发表18篇论文（2009年录用了8篇），在分布式领域的旗帜会议IEEE ICDCS上发表12篇论文，在包括ACM Multimedia, mobihoc,IEEE ICNP、IWQoS等在内的其他本领域重要会议上发表60余篇高档次学术论文。在16个国际学术会议上受邀做报告，多篇论文获得最佳会议论文，并获IEEE通信学会Leonard G. Abraham论文奖（IEEE JSAC年度最佳论文）。
演讲题目：Serving P2P Streaming Systems with GPU-accelerated Network Coding
Abstract: Peer-to-peer (P2P) streaming has recently received much research attention, with successful commercial systems showing its viability in the Internet. In the first part of this talk, we present a new set of design principles towards the design of new streaming protocols, with the help of network coding. These new streaming protocols enjoy short initial buffering delays, high playback qualities, resilience to peer dynamics, and smaller bandwidth costs on dedicated streaming servers. Even though network coding is beneficial, high computational complexity has limited applications of network coding. In the second part of this talk, we explore the computational limits of parallel network coding on modern CPUs. We show that Graphics Processing Units (GPUs) are a prime alternative to CPUs on streaming servers. We present our recent work on a highly-optimized implementation of GPU-accelerated network coding, offering significant coding gains at a much better price/performance ratio.
