Engineering performance using control theory
Joseph L. Hellerstein, Yixin Diao, et al.
CMG 2007
Over the last few years, there has been considerable success with applying control theory to computing systems. Our experience has been that there are several commonly occurring control problems in computing systems - translating between service oriented units (e.g., response times) and effector (actuator) units (e.g., the maximum number of connected users); optimizing resource usage; regulating service levels to enforce service level agreements; and adapting to disturbances such as changes in workloads. Developing control systems that address these problems involves challenges related to modeling the managed element (plant); handing sensor data that are noisy, incomplete, and inconsistent; dealing with effectors that have complex effects that often do not correspond well to the control objectives; and designing control systems (especially filters, the choice of measured outputs, and time delays).
Joseph L. Hellerstein, Yixin Diao, et al.
CMG 2007
Sheng Ma, Joseph L. Hellerstein
ICDM 2001
Yixin Diao, Joseph L. Hellerstein, et al.
ACC 2004
Joseph L. Hellerstein
IEEE Transactions on Knowledge and Data Engineering