SymPerf: Predicting Network Function Performance

 1@inproceedings{rathSymPerfPredictingNetwork2017,
 2  author = {Rath, Felix and Krude, Johannes and Rüth, Jan and Schemmel, Daniel and Hohlfeld, Oliver and Link, Jó Ágila Bitsch and Wehrle, Klaus},
 3  title = {{{SymPerf}}: {{Predicting}} {{Network}} {{Function}} {{Performance}}},
 4  booktitle = {{{SIGCOMM}} {{Posters}} and {{Demos}} {{(SIGCOMM}} {{Poster}} 2017)},
 5  location = {Los Angeles, CA, USA},
 6  pages = {34--36},
 7  year = {2017},
 8  month = {aug},
 9  day = {22},
10  doi = {10.1145/3123878.3131977},
11}

The softwarization of networks provides a new degree of flexibility in network operation but its software components can result in unexpected runtime performance and erratic network behavior. This challenges the deployment of flexible software functions in performance critical (core) networks. To address this challenge, we present a methodology enabling the prediction of runtime performance and testing of functional behavior of Network Functions. Unlike traditional performance evaluation, e.g., testbed testing or simulation, our methodology can characterize the Network Function performance for any possible workload only by code analysis.