Embedded Files
SANDIE: SDN-Assisted NDN for Data Intensive Experiments
2016-PresentNortheastern University, Boston, MAJoint with CERN (European Organization for Nuclear Research), California Institute of Technology and Colorado State UniversitySANDIE: SDN-Assisted NDN for Data Intensive Experiments
The SANDIE project aims to meet challenges facing data-intensive science programs such as the Large Hadron Collider (LHC) high energy physics program at CERN, based on VIP joint caching, forwarding and congestion control framework in the novel Named Data Networking (NDN) architecture.
Photo on the right: Prof. Edmund Yeh (my advisor) giving a presentation on SANDIE demo at SC '19
- Data analysis on 2 years of global worklog on CMS (the experiment that found Higgs-Boson) workflow patterns;
- Verified excellent performance gain achieved by VIP framework on fetching CMS datasets by simulations with a self-developed packet-level JAVA simulator;
- Coordinated and participated in two parallel developments of VIP framework, based on NDN-DPDK forwarder (https://www.dpdk.org/) by NIST and NDN-Forwarding-Daemon (NFD, http://named-data.net/doc/NFD/) by NDN community;
- Led and coordinated the configuration of a continental testbed, involving ~ 10 parties regarding regional VLANs and testbed machines;
- Designed, coordinated and successfully accomplished the project demonstration at SuperComputing 19 (Denver, CO).
Proactive Caching for Low-Access Delay Services under Uncertain Predictions
2018-Present Northeastern University, Boston, MA Advisor: Prof. Edmund Yeh (Northeastern University), Prof. Atilla Eryilmaz (Ohio State University)Proactive Caching for Low-Access Delay Services under Uncertain Predictions
- Derived a tight upper bound on how much proactive service per request that the system can support for general proactive caching strategies under uncertain predictions;
- Analyzed the asymptotic behavior of a family of threshold-based proactive strategies with a novel Markov chain.
- Proved optimality when choosing threshold as the upper bound derived previously;
- Proved the most commonly used First-Come-First-Serve proactive strategy always has worse delay than the optimal threshold-based strategy in all the non-trivial cases.
- Carried out extensive numerical experiments to verify the results. Designed a modified version of the optimal threshold-based strategy which works well with smaller prediction window size.
Figure above: Queue Evolution Comparisons under the most commonly used Earliest-Deadline-First (EDF, Orange) proactive strategy and proposed Optimal-Threshold strategy (UNIFORM , Green), with reactive scheme (Blue) as a baseline.
VIP: A Framework for Joint Dynamic Forwarding, Caching and Congestion Control in Named Data Networks
2013-2016 Northeastern University, Boston, MA Advisor: Prof. Edmund Yeh (Northeastern University)VIP: A Framework for Joint Dynamic Forwarding, Caching and Congestion Control in Named Data Networks
- Implemented and optimized a packet-level simulator for VIP caching, forwarding & congestion control algorithms;
- Designed and implemented VIP chunk-level strategy with minimal impact on performance;
- Designed and tested a proactive VIP caching algorithm. Effectively increased convergence speed;
- Implemented a 3D animated visualization tool which demonstrated the insights of VIP framework.
- Studied performance of VIP framework in practical scenarios including core network, data center network, wireless edge network and specifically High-Energy-Physics network;
Videos above: Visualization of VIP algorithm on a grid topology. Bar-levels reflect queue sizes of requests for each data object. Upper, left and right corners are requesters, and lower corner is the data source. Red-dotted bar indicates that this object is being cached. The downward gradients from the requesters to the source implies the insight of the backpressure algorithm in control plane.
NDN-Based Adaptive Video Streaming System
2018-now Northeastern University, Boston, MA Advisor: Prof. Edmund Yeh (Northeastern University)NDN-Based Adaptive Video Streaming System
- Studying adaptive video streaming formats (e.g. MPEG-DASH) and augmented video coding for multi-resolutional video streaming;
- Incorporated VIP framework in NFD and tested it with NDN video applications;
- Created a video library of ~1000 videos of 4 level of resolutions (720p, 1080p, 2k, 4k) for the purpose of test.
Other Projects
Other Projects
Throughput Optimal Scheduling in Multimedia Sensor Networks with High Traffic Load
course project for 'Special Topic: Wireless Sensor Networks', 2016 Northeastern University, Boston, MA with Khashayar Kamran- Studied backpressure algorithm in a wireless multimedia sensor network scenario;
- Simulated on a random wireless sensor network topology, and showed excellent performance achieved by backpressure algorithm.
Survey: Evolution from 0G to 4G
course project for 'Mobile and Wireless Networking', 2014 Northeastern University, Boston, MAApplications of Mathematics and Control Theories in Wireless Network Research
Bachelor Thesis , 2013 Shanghai Jiao Tong University, Shanghai, ChinaAdvised by Prof. Xinbing WangHotel Room Management System
course project for 'Software Engineering' I and II, 2009-2010 Shanghai Jiao Tong University, Shanghai, China- Led a group of 5 members in planning, designing, documenting and developing;
- Developed a hotel room management system with C++ with Visual UI, with complete documentations.
Page updated
Google Sites
Report abuse