Environment-Aware Work Load Prediction in Edge Computing

Computing resources can be brought close to users by mobile edge computing technologies, which can meet users’ low latency and computing needs. To address the work load imbalance generated by user movement, most methods focus on time-series prediction. However, no one considers environment information. Based on this, a work load prediction method considering environment information is proposed. The ET-GCN model consists of a graph convolutional network and a gated recurrent unit network, in which the geographic regions covered by the edge servers or base stations are abstracted as nodes in the graph. Then the GCN learns the geospatial features of the city, and the gated recurrent units capture the temporal dependencies. Experiments show that the RMSE of ET-GCN model is 4 % less than the existing baseline algorithms.