Information collection architecture for performance measurement of computer networks

A new Information Collection Architecture (ICA) is presented that solves the data collection problem for real-time performance measurement of high speed computer communication networks. Despite some claims to the contrary, traditional data collection techniques either overflow available disk storage or exceed the real-time processing capabilities of embedded processors, and hence, they fail to support useful or accurate real-time analysis for network management. The ICA overcomes these deficiencies with a unique hardware/software implementation based on Finite State Machines (FSMs).; We present a set of procedures for using the ICA to provide performance measurement for communication protocols. We also extend previous performance measures to include additional communications protocols.; We provide a formal use of regular expressions to specify recognition sequences of interest for communication protocols. These recognition sequences correspond to the parameters used by performance models. We investigate alternative FSM structures for recognizing these regular expressions. We introduce a structure based on predecessor and successor component FSMs that allows greater implementation flexibility than the alternative structures.; A prototype ICA data collection device has been developed and built to demonstrate practicality, feasibility, and utility of the ICA approach on 16 Mbps and 4 Mbps token ring local computer networks. We discuss the design and implementation of this device for a given set of performance indicators.; We provide a formal definition of a time transformed FSM and show how to use this FSM for performance parameter recognition in high speed networks. We propose an ICA implementation architecture for real-time performance measurement in 1 Gigabit per second networks using presently available technology.; We use the ICA prototype device to measure performance of operating token ring networks, and we compare these measurements with measurements taken by a traditional network performance monitor. Measured data from both approaches are also compared with simulation results. These comparisons show that the ICA accomplishes accurate performance measurement for real-time network management by collecting only a small fraction of the data required by traditional approaches.