Past and Current Projects
Understanding the Performance of Complex Computer Systems
Modern computer systems have become so complex, with so many layers of hardware and software, that computer systems researchers have difficulty understanding the performance of their systems. For example, a program may slow down during execution, and this could be because of the structure of the hardware, the operating system, the virtual machine on which the program is running, or the program itself. Consequently, computer systems researchers have become experimentalists, instrumenting their programs to read out various sorts of data as the programs run. The situation is much like that in cognitive science: cognitive scientists want to understand the performance of a complex system (the brain), and collect various sorts of data from the system to draw inferences about the underlying mechanisms. We are using the techniques of computational cognitive modeling to understand the performance of complex computer systems, relaying primarily on statistical machine learning models.
Students
Le Huang
Scott Richardson
Michael Otte
Collaborators
Amer Diwan (Computer Science, Colorado)
Matthias Hauswirth (Lugano)
Peter Sweeney (IBM Yorktown Heights)