Data Sharing-Aware Online Algorithms for Task Allocation in Edge Computing Systems

Many of the tasks offloaded to edge devices perform computation to analyze sensing data. Transferring this data from end-user devices to edge servers leads to increased latency and congestion in the edge network. Since many of the offloaded tasks may require processing the same data items, the task allocation algorithms can exploit this to reduce the traffic in the networks and the number of edge servers needed to execute the tasks. Therefore, in this paper we design online algorithms for task allocation in edge computing systems that take into account the sharing of data among the tasks offloaded to the same server. We perform an extensive performance analysis by comparing our proposed algorithms with several sharing-oblivious baseline algorithms. The results show that our algorithms are able to reduce the amount of data transferred in the network by $30.2 \%$ to $\mathbf{9 2. 8 \%}$ and the number of utilized servers by $\mathbf{1 \%}$ to $82.8 \%$ compared to the sharing-oblivious baseline algorithms.