-->
My Photo

Sung-Han Lin (林嵩翰)

Ph.D. candidate
Department of Computer Science
University of Southern California

Office: Henry Salvatori Computer Science Center (SAL), Room 200
Address: 941 Bloom Walk, Los Angeles, CA 90089, USA
Email: sunghan [at] usc.edu

Research

I am working in Quantitative Evaluation & Design Research Group with Prof. Leana Golubchik. I received my B.S. degree in Computer Sience and Information Engineering, National Taiwan University in 2006, and M.S. degree from Communication and Multimedia Laboratory in Computer Science and Information Engineering, National Taiwan University in 2008.

I am interested in Computer Networking, especially Distributed Systems and P2P Networks, Performance Modelling and Analysis, Economics and Game Theory, and Cloud Computing.

Professional Experience

Publications

Projects

Improve the performance of SQL-MR Execution Engine - Teradata Labs Intern

I profiled the performance of SQL-MapReduce Execution Engine and re-designed the mechanism to reduce the data transmission and I/O time between user defined functions and databases.
(1) Aggregating the output: Instead of sending result row by row in the original design, I caches rows and send it once to reduce the I/O time. This improves the performance by at least 20%.
(2) Eager sending: Sending cached data immediately if it is available. This improves the performance by 10%.

MediaWall Framework - FXPAL Intern

I designed and developed media wall framework which instantiates and controls virtual machines to enable diversified screen presentations that are not limited by pre-installed projectors. Since this system works in a network, I built a centralized system to manage network resources and handle interactions between remote machines and services to enhance the performance. Moreover, I developed APIs to provide programmatic access from web-enabled platforms. To provide access interface, I programmed Web-based Graphical User Interfaces to enable presenters to manage projected screens on walls

Speed up loading large data sets to a Facebook-like system on Cassandra

Built Cassandra clusters with small-scale OpenStack virtual machines suitable for Facebook-like social network. Designed mechanisms equally distributing the workload to all running virtual machines. Coded multi-thread systems to handle different kinds of inputs without leaving the system idle.

Implemented a distributed large-scale Digital Signage system.

Implemented a distributed large-scale Digital Signage system. Developed partial storage systems for video contents and advertisements by combining Content Distribution Network and Peer-to-Peer technologies

Implemented a real P2P IPTV system

Coded P2P Internet Protocol Television(IPTV) systems supporting channel browsing in Windows systems. Explored and measured effect of parameter settings for different network capability environments.

Performance Analysis for the Speed-Sensitive Channel Assignment

Developed and simulated probability models to correctly analyse the performance of channel assignment with rapid cell phone hand-offs. Explored the effect of adopting random walk or human walk to probability models for assigning channels.

Exploit File Similarity in Peer-to-Peer Networks

Explored probability of having the same chunks among files in eMule file-sharing environments. Developed sharing mechanism capable of using similar chunks to speed up peer-to-peer file downloading.

Financial Expert System

Designed an Expert System to evaluate a company’s financial position.