A Research of Compute First Networking Based on Information-Centric Networking Cache Mechanism
Qianwen Xu, Shuangyin Ren, Jingchao Wang, Zhongbo Li, Wei Gao · 2025
Compute first networking is a new network model that deeply integrates distributed computing resources with network communication technology, aiming to realize global optimization of computing resources through intelligent scheduling. However, the existing computing and network convergence architecture based on traditional TCP/IP networks, centralized resource management has shortcomings in scheduling flexibility and scalability, and information-centered network (ICN) has the potential to solve these problems. The ICN uses content names rather than IP addresses as the core for routing and transmission, and supports caching, multipath distribution, and dynamic resource discovery. Its decentralized structure meets the requirements of distributed resource pooling in Compute first networking, providing a good environment for flexible routing scheduling. To solve this problem, a new computing resource scheduling architecture ICCRSA based on ICN cache mechanism is proposed. In this architecture, a caching mechanism based hash multicast tree power routing algorithm HMCR is proposed. The power node is mapped to the compute resource node through the hash function, and the compute request is directly mapped to the compute resource node with the same hash value. Each resource location node maintains the state of the recently visited compute resource. Otherwise, a multicast distribution tree containing resource location nodes and requesters is constructed to transmit the state of available computing power information in parallel, and cache along the path based on node location and resource popularity. The experimental results show that compared with other caching schemes, this architecture has obvious advantages in enhancing scheduling flexibility and improving scheduling success rate.