The complexity of today's networks and distributed applications exceeds our ability to measure and manage these systems. Control Room has been designed as a testbed for applying feedback control algorithms to software systems to provide improved and more predictable performance. Originally designed as a centralised system that provided a database of performance information for control decisions the architecture has recently been redesigned to support distributed collections of data.
Control Room's new architecture features distributed repositories of data, a publish and subscribe paradigm and applets as information caches. The new architecture allows control stations to dynamically adjust the data they monitor and the way in which it is presented. It has allowed us to experiment with the use of high level objects such as graphs as a form of data transport.
|
| John H. Hine |