Predicting Distributions of Service Metrics using Neural Networks
2018 (English)In: 14th International Conference on Network and Service Management, CNSM 2018 and Workshops, 1st International Workshop on High-Precision Networks Operations and Control, HiPNet 2018 and 1st Workshop on Segment Routing and Service Function Chaining, SR+SFC 2018, 2018, p. 45-53Conference paper, Published paper (Refereed)
Abstract [en]
We predict the conditional distributions of service metrics, such as response time or frame rate, from infrastructure measurements in a cloud environment. From such distributions, key statistics of the service metrics, including mean, variance, or percentiles can be computed, which are essential for predicting SLA conformance or enabling service assurance. We model the distributions as Gaussian mixtures, whose parameters we predict using mixture density networks, a class of neural networks. We apply the method to a VoD service and a KV store running on our lab testbed. The results validate the effectiveness of the method when applied to operational data. In the case of predicting the mean of the frame rate or response time, the accuracy matches that of random forest, a baseline model.
Place, publisher, year, edition, pages
2018. p. 45-53
Keywords [en]
Generative Models, Machine Learning, Network Management, Service Engineering, Decision trees, Learning systems, Routing algorithms, Baseline models, Cloud environments, Conditional distribution, Gaussian mixtures, Generative model, Operational data, Service assurance, Forecasting
National Category
Natural Sciences
Identifiers
URN: urn:nbn:se:ri:diva-37760Scopus ID: 2-s2.0-85060906697ISBN: 9783903176140 (print)OAI: oai:DiVA.org:ri-37760DiVA, id: diva2:1287655
Conference
14th International Conference on Network and Service Management, CNSM 2018 and Workshops, 1st International Workshop on High-Precision Networks Operations and Control, HiPNet 2018 and 1st Workshop on Segment Routing and Service Function Chaining, SR+SFC 2018, 5 November 2018 through 9 November 2018
Note
Funding details: VINNOVA; Funding text 1: The authors are grateful to Erik Ylipää with RISE SICS, as well as to Andreas Johnsson, Farnaz Moradi, Christofer Flinta, and Jawaad Ahmed with Ericsson Research for fruitful discussion around this work. This research has been partially supported by the Swedish Governmental Agency for Innovation Systems, VINNOVA, through project SENDATE-EXTEND.
2019-02-112019-02-112021-11-26Bibliographically approved