With the rollout of the 5G network around the globe, a massive number of edge servers have been deployed to host online applications demanding low service latency for users. These edge servers constitute multi-access edge computing (MEC) systems. Running 24/7, edge servers consume tremendous energy and take up a great part of global carbon emissions. The edge energy-saving (EES) problem is needed to facilitate energy-efficient edge resource provisions. Unfortunately, existing energy-saving approaches designed for data centers are becoming impractical. First, edge servers are used to provide services to a specific geographical area. Its energy utilization is impacted by the temporal distribution of users within its coverage. Second, a user could be accommodated by any of its neighbor edge servers. Third, it is possible to activate and deactivate individual physical machines that facilitate an edge server as needed. Thus, EES is designed to save the system energy of physical machines in a long term by serving the users over time. EES problem has been formulated systematically and its problem hardness has been analyzed theoretically, then we propose EESaver (Edge Energy Saver) for formulating EES strategies dynamically over time to facilitate green MEC. EESaver's superior performance is tested comprehensively.
原文链接
分享
引用
