Software and Programmer Efficiency Research Group

We are interested in analyzing and visualizing system behavior. In particular, we study visualization and analysis approaches and tools that enable the efficient and effective interactive exploration of system performance traces. We focus on finding approaches and tools for "performance understanding", for example by identifying causal relationships between phenomena to explain the reasons of performance problems.

So far, our efforts have focused on the following two topics:

• Trace alignment enables reasoning across traces of performance metrics collected in multiple executions.
• Trace visualization enables the interactive analysis of dynamic system behavior and performance.

Publications

• SP&E. TraceAnalyzer: A system for processing performance traces
• ISPASS10. LagAlyzer: A latency profile analysis and visualization tool
• MICRO'07. Time Interpolation: So Many Metrics, So Few Registers
  (Comparing trace alignment to traditional performance counter multiplexing)
• CC'06 demo. Demonstration: On-Line Visualization and Analysis of Real-Time Systems with TuningFork
• STMCS'06. Understanding Performance of Multi-Core Systems using Trace-Based Visualization
• NFS/NGS @ IPDPS'06. Aligning traces for performance evaluation
• OOPSLA'05. Automating Vertical Profiling
  (Using dynamic time warping to align performance metric time series)
• EMSOFT'05. High-level real-time programming in Java
  (TuningFork for online visualization of real-time VM/GC traces)
• OOPSLA'04 demo. Performance explorer: understanding java application behavior
• PAC2'04. Performance and Environment Monitoring for Whole-System Characterization and Optimization
  (Performance Explorer for offline visualization of PEM traces)