Internal Task Offloading Deployment Mechanism of In-Network Computing Devices for Holographic-Type Communication

As a novel service, holographic-type communication (HTC) involves processes such as content generation, compression-decompression, security policy, and rendering display. This requires not only significant computational power from the network but also the ability to meet the latency requirements of holographic services. To address the high computational power and low-latency transmission needs of holographic services, this paper proposes an optimized task offloading mechanism for in-network computing devices. It accelerates holographic communication computational tasks by deploying portions of these tasks to network devices with computational capabilities. Through specialized hardware acceleration, the mechanism optimizes the overall latency. Consequently, this paper designs a task allocation algorithm based on reinforcement learning, which distributes tasks to specialized hardware under latency constraints, thereby increasing the number of tasks deployed at in-network computing nodes. The experiments indicate that the proposed mechanism can increase the number of tasks deployed. Compared with the baseline algorithm, the number of computational tasks handled by a single in-network computing node is increased by \(20\%\) .