Statement of Purpose for Stanford University v2
I want to pursue my master’s degree in computer science with focus on computer networking.
During my undergraduate studies in Elctronic and Information Engineering Major at Zhejiang
University, I gained solid analytical and programming skills. I ranked the 3 rd in 130 students in
my third semester. I got a GPA of 92.2 (100-scale) in all mathematic courses and gained 96 in
C-programming.
I entered the Department of Electronic Engineering department at Tsinghua University to further
my graduate study. In the courses, I not only grasped the fundamental theories but also gained
insight of the cutting edge technologies in communication and networking. The stochastic process
and queuing theory are powerful mathematic models to predict network performance in dynamic.
For example, the flash-crowd and peer heterogeneity in P2P network can be simulated with the
On-Off process. Besides, I realized that the degree distribution and clustering coefficient in
Complex Network give insight of the network resilience and connectivity. The human contact
pattern of online and mobile social network will guide the mapping of information network into
human social network. In my favourite course “Advanced Network Technologies”, I surveyed the
“Network Coordinates”, and found the mapping between network latency and geometric distance
suffers from the TIV phenomenon because in current Internet architecture source and destination
addresses cannot uniquely define a flow (route) between them.
这一段有点多，写三件事就行
At the same time, as a research assitant, I started my research work in the Lab of New Generation
Network. In the SIP and IPv6 based P2P multimedia project, I implemented the SIP protocol as an
accountable middleware, to separate the control signaling and the data flow. Besides, I found that
the disconnectivity between IPv4 and IPv6 hosts will bring heavy joining failure and latency
deterioration to structured overlay, especially the multicast tree. I modeled the acception capacity
of the IPv4, IPv6, and dual stack nodes into a three-dimension markov process, and calculated the
joining failure with different nodes distribution. And I developed my heauristic algorithm to
eliminate the joing failure and latency deterioration. With this practice, I realized that the
dis-match between overlay and underlining network is the main reason of cross-ISP traffic
problem and overlay performance degradation.. The locality approach like P4P aims to build
baised overlay to match the network.
In the project of On-demand web service composition prototype platform, I lead my group to
develop mutli-media web services for Travelling. To enhance the system compatibility and
flexibility, I designed a unified Interface for cross Web Service Interaction. And I was pretty
enjoying the project, as I can try out new ideas in it. I proposed and implemented the Short
Message Search service as many user scenarios of our platform are in the mobile environment,
which prefer low-cost short message media for information exchange.
My internship in IBM CDL was mainly on web development and product test. This experience
gained me insight of software development and product test.
要么上面两段合并一下吧
I hope to study the Computer Networking for master’degree with focus on two directions:
1. P2P streaming with heterogeneous peer capacity. Current research either aims to optimize
overlay by capacity peer clustering or leveraging special coding scheme like Multiple Description
Codec or Multi Layer Codec. The content-aware scheduling approach proposed by Professor
Benrd Girod’s group has natural quality adaptive property by media content prioritization and is
compatible with current media codec. However, the tree based overlay suffers from network churn
and.？？ With my experience in overlay construction and content scheduling, I want to study
the optimized ？？
2. Programmable Network. Professor MeKeown and the OpenFlow project team aim to build
pragrammable router and switch network with fine-grained flow table and centralized controller.
This approach provide flexibility, controllability and isolation for new protocol and algorithm
experiments on the layer2/layer3. However, the centralized controller suffers from single point of
error and will be the bottleneck when network scales. I want to develop a distributed controller
overlay and developed load-balance scheme.
The Stanford University’s flagship “Clean-Slate Internet” project drives me to apply to the
Stanford University. The mobile networking drives the innovation and revolution of current
Internet architecutre. And I would like to expose myself to these research projects, such as
OpenFlow, either help to build the Pragrammable Network testbed, or develop innovative
algorithms on media streaming over this testbeds. The close relationship with industry is another
reason I choose Stanford. I like to gain more experience of developing the real systems in
computer networking with research and industries values.