Evaluation of Internal Delay Inference in Queuing Networks



Stoner, David E

Journal Title

Journal ISSN

Volume Title



Statistical inference of internal computer network characteristics using only externally made measurements is extremely useful in the analysis of highly complex networks. This the- sis seeks to implement and test an expectation-maximization (EM) algorithm that uses these observations to estimate total end-to-end network delay density, link delay density and prob- ability of route selection. The EM algorithm in question was tested using source/destination delays generated from a custom queuing network simulator. The parameters of the queuing network were varied in order to determine the e ectiveness of the algorithm on Jackson-type networks as well as more realistic networks. The subsequent results of the algorithm are compared against the actual network simulation data to evaluate the performance of the algorithm.



Bivariate Markov chain, EM algorithm, Queuing network, Python