Elastic Provisioning of Network and Computing Resources at the Edge for IoT Services

The fast growth of Internet-connected embedded devices demands new system capabilities at the network edge, such as provisioning local data services on both limited network and computational resources. The current contribution addresses the previous problem by enhancing the usage of scarce edge resources. It designs, deploys, and tests a new solution that incorporates the positive functional advantages offered by software-defined networking (SDN), network function virtualization (NFV), and fog computing (FC). Our proposal autonomously activates or deactivates embedded virtualized resources, in response to clients' requests for edge services. Complementing existing literature, the obtained results from extensive tests on our programmable proposal show the superior performance of the proposed elastic edge resource provisioning algorithm, which also assumes an SDN controller with proactive OpenFlow behavior. According to our results, the maximum flow rate for the proactive controller is 15% higher; the maximum delay is 83% smaller; and the loss is 20% smaller compared to when the non-proactive controller is in operation. This improvement in flow quality is complemented by a reduction in control channel workload. The controller also records the time duration of each edge service session, which can enable the accounting of used resources per session.

Automated summary

This study proposes a solution that incorporates software-defined networking (SDN), network function virtualization (NFV), and fog computing (FC) to autonomously activate or deactivate embedded virtualized resources in response to clients' requests for edge services. Extensive tests show the superior performance of the proposed elastic edge resource provisioning algorithm, resulting in higher flow rate, smaller delay, and reduced loss compared to non-proactive controller operation. Furthermore, the controller records the duration of each edge service session, enabling resource accounting per session.