Dynamic Multi-Tier Resource Allocation Framework for Metaverse

Since Metaverse requires enormous resources that have never been seen before, resource management is one of the key challenges hindering its widespread deployment and performance. Thus, this paper introduces a novel framework that can effectively and intelligently manage various types of resources in different network layers to meet strict requirements of Metaverse, e.g., terabit-per-second-level bitrate and millisecond-level latency. In particular, this framework is built based on a multi-tier resource allocation framework and two innovative techniques: (i) MetaSlice decomposition providing a flexible and effective solution in deploying, managing, and updating Metaverse applications, and (ii) MetaInstance maximizing resource utilization by exploiting similarities among Metaverse applications. Moreover, to address the dynamic, uncertain, and real-time resource demand in Metaverse, we develop an intelligent algorithm that can quickly find the optimal resource allocation for the system. The key idea of this algorithm is to automatically learn the optimal policy through interactions with the environment without requiring complete information of the environment, which is infeasible to be obtained in Metaverse. The simulation results show that the proposed framework can not only improve the long-term revenue for the Metaverse provider up to 1.8 times but also enhance user experience (e.g., request acceptance rate) near 1.5 times compared with other baseline schemes.