Energy-Efficient Distributed Learning and Sharding Blockchain for Sustainable Metaverse
Peng Wang, Lan Wei, Wen Sun, Haibin Zhang, Yan Zhang
Abstract
Nowadays, researchers have started to conceptualize Metaverse with the vision of constituting a fully immersive, hyper spatiotemporal, and persistent interconnected virtualized world. Such network evolution poses sustainability concerns due to its enabling technologies, such as compute-intensive Artificial Intelligence (AI) and energy-consuming blockchain. Combining distributed learning and blockchain shows great potential to solve the energy efficiency issues in Metaverse through secure resource scheduling and decentralization of computing. However, with the expansion of the Metaverse scale, the increased energy consumption and storage of blockchain are still intolerable. Sharding blockchain becomes a feasible solution to efficiently improve energy efficiency and scalability by dividing blockchain into multiple smaller groups called shards. Toward this end, we have proposed a sustainable Metaverse architecture, combining distributed learning and sharding blockchain to tackle the energy efficiency challenges. The proposed sharding mechanism with incentive achieves the parallelization of computing and storage in Metaverse, while guaranteeing the security and activity of distributed learning. Numerical results show that the proposed framework improves energy efficiency and eases pressure on data storage.