Electrical Engineering Department
University of Southern California
Advisor: Professor Leana Golubchik
Email Address: yuanyao AT usc DOT edu
Performance Modeling and Analysis
Use of mathematical and algorithmic methods in performance modeling and analysis of state of the
art computer systems and applications, such as peer-to-peer systems, wireless sensor systems,
and data centers.
Performance Oriented Design
Understanding of underlying performance trade-offs and performance-oriented design of simple
yet effective algorithms for computer systems and distributed applications.
Power Cost Reduction in Large Scale Distributed Computing
In this project we try to reduce power cost of large scale distributed computing failities, while
maintaining quality of service. We leverage both temporal and spatial differenes of power prices
and workload and formulate it as an stochastic optimization problem. We provide distributed
solutions to this problem and show that it can result in significant power cost savings. Anomaly Detection in Sensor Systems
In this project we develop automated algorithms for identifying interesting events(termed
anomalies) in sensor data. The algorithms we develop are able to accurately detect short
term and long term anomalies without prior knowledge of anomaly characteristics. These
algorithms are also robust to sensor faults and operate in (near) real time.
Peer-to-peer Streaming Systems
In this project we study performance trade-offs in structured peer-to-peer intended for streaming
application. We design streaming algorithms based on multi-tree and hypercube structures.
These algorithms achieve provable quality-of-service guarantees, and specifically hiccup free
video streaming with low start-up delay and low buffer space requirements.