Ju Wang
	Ph.D. Candidate Department of Computer Sci. & Engineering University of California, San Diego jwang @ cs.ucsd.edu 858.534.5486 (Office)	Resume Doctoral Dissertation Project Publications and Talks Research Interest My research interests are in system and network area, primarily on security and fault tolerance. In the past, I have also been involved in several interesting projects on distributed components models, binary instrumentation, high-performance communication systems and so on. q Denial of Service Attacks My doctoral project is to explore the use of overlay networks to protect applications from Denial-of-Service (DoS) attacks. A high level description can be found here. Publications and talks on this topic can be found here. q Overlay Network Performance A side project Andrew and I did in summer 2003. The goal is to understand the deciding factors for the overlay network performance. The overlay network topology, the underlying network (e.g. IP network) topology, and the mapping between overlay nodes and hosts in the underlying network are the three key factors. From limited case study of the underlying network being a sub-graph of the Internet, we found that the achievable performance of overlays is primarily decided by the mapping. The topology of the overlay has no significant impact as long as there is sufficient connectivity in the overlay topology. However, graphs such as de Bruijn and hypercube, which are well studied in the past, have shown signs of supremacy over the recently popular overlays such as Chord. In particular, for certain underlying topologies, de Bruijn and hypercube significantly outperforms Chord. Whether this holds for Internet graphs in general is still unknown. Due to time constraint, this project is currently on hold. Hopefully it will be resumed in the near future. q Fault Tolerance: In the context of the Agile Objects project in CSAG, I did an extensive survey on fault tolerance research, particularly on object groups. These projects are primarily based on the state machine model and apply group communication protocols in the context of distributed objects. Here is a summary of this survey. Related Research Projects on Fault Tolerant Objects q Binary Instrumentation I spent almost two years working in Entropia (www.entropia.com), a startup company on desktop Grid systems. I was a core member of the VM team and a primary contributor to the design and implementation of Entropia © Virtual Machine and our binary instrumentation technology. Furthermore, I successfully ported a large range of windows applications to our VM platform in their binary form, including 3DSMax© and Microsoft Office© applications. J q Distributed Components In my Master’s project, I built a runtime system to transparently replicate COM (Component Object Model) objects in unmodified COM applications. The key is to transparently intercept all the COM instantiations and interface invocations. Our approach is to put an interceptor inside the COM-RPC layer, so that all the calls can be intercepted without any modification to the application. We successfully ported unmodified third-party COM applications to our runtime system. Details about this project can be found in the Thesis. Mater’s Thesis: “Transparent Replications for Component-based Applications”. q High-Performance Communication System for Clusters A project I was involved in back in Tsinghua. We built a Fast Message Passing communication system for a Multi-Processor SUN SPARC-II cluster connected with Myrinet. I built the local communication subsystem on shared memory. Our work was published on the journal of Parallel and Distributed Computing Practices, vol 1. no 4. 1998. Resources Slides from UCSB Network Security Class A few AI gaming software I made with my friends (Zhou Lin and Wan Min) back in Tsinghua. Gomoku (zip) and Chinese Checker (zip). My friends' links: Xianhan Zhu