On-Demand Allocation of Cryptographic Computing Resource with Load Prediction

Xiaogang Cao, Fenghua Li, Kui Geng, Yingke Xie, Wenlong Kou

Abstract: Cryptography-as-a-Service” provides convenience for users to request cryptographic computing resources according to their needs. However, it also brings challenges for resource management, such as the constantly changing load, large numbers of users, and complex resource topologies. To address those issues, this paper proposes a load-predicted-based resource allocation algorithm for cryptographic computing resources. Firstly, we propose a load-based cryptographic computing resource allocation model that can clearly describe the dynamic status of resources. Then, we design a load predictor using time series analysis and a random forest model, which can quickly predict the load of cryptography service requests during service time. Finally, we develop a load-predicted-based greedy algorithm for cryptographic computing resource allocation. Experimental results show that energy consumption is reduced by about 20% at most compared to the baseline allocation algorithm.