Performance Evaluation of Some Adaptive Task Allocation Algorithms for Fog Networks

Ioanna Stypsanelli (1) , Olivier Brun (1) , Balakrishna Prabhu (1)

1 LAAS-SARA - Équipe Services et Architectures pour Réseaux Avancés

The Fog Computing paradigm provides a seamless bridge between the Cloud and the Edge computing architectures. Depending on the their QoS requirements, tasks can be processed in either the Edge or the Cloud or migrated from one to the other. For the benefits of this flexible architecture to be seen, task allocation algorithms should be able to adapt to the load in the Fog and in the Cloud, and send the tasks to a lightly loaded resource. Current task allocation algorithms in Fog computing literature use a simple offloading strategy: a task is first sent to the nearest Fog node. If the nearest Fog node is saturated, it offloads the task to the Cloud. Although simple to implement, such a strategy disregards available resources in the Fog nodes that could be further away but less congested. Using a discrete-event simulation approach which relies on the network simulation framework OMNeT++, we show that simple adaptive algorithms based on congestion estimation outperform the standard nearest node algorithm. For this, we choose four distributed routing algorithms that were proposed for other networking applications but which are also well-suited for Fog networks. These algorithms improve the resource usage as well as reduce the mean job processing times in scenarios with offloading as well as without offloading. Our usecase is that of Fog networks that use the cellular network: base stations (access nodes) forward traffic to computing nodes (Fog and Cloud nodes) in a distributed way without coordination and sharing of state-information. We evaluate the performance of these algorithms on several scenarios that include sudden changes in the arrival rate of requests (to model peak hours) and changing the variance of request processing times (to study the robustness to request-size distributions) to understand the advantages and drawbacks of each of them.

Mots clés: adaptive algorithms; OMNeT++; multi-armed bandits; task allocation; offloading; Base stations; Adaptation models; Heuristic algorithms; Computational modeling; Routing; User experience; Resource management

Domaines: Réseaux et télécommunications [cs.NI]

Ioanna Stypsanelli, Olivier Brun, Balakrishna Prabhu. Performance Evaluation of Some Adaptive Task Allocation Algorithms for Fog Networks. IEEE 5th International Conference on Fog and Edge Computing (ICFEC 2021), May 2021, Melbourne, Australia. ⟨10.1109/ICFEC51620.2021.00020⟩. ⟨hal-02972802⟩