As a key component of next-generation communication technologies, Holographic-type Communication (HTC) places stringent demands on the network, necessitating low latency, high bandwidth, and substantial computing power. The requirements create considerable challenges for the dynamic coordinated scheduling of network resources. Existing resource scheduling approaches primarily focus on edge computing, and employ distributed computational deployments to reduce end-to-end latency. However, the limited computing resources of edge computing present a significant bottleneck in HTC. With the evolution of the In-Network Computing (INC) paradigm, intelligent network-layer components, exemplified by Data Processing Units (DPUs), have demonstrated advantages such as data plane acceleration, offering promising support for HTC over edge computing. Nevertheless, the problem of dimensional resource scheduling optimization within INC frameworks is a critical research challenge. In this research, we propose an INC-based multidimensional resource model, customized for three representative holographic services. The model integrates transmission, computing, and caching resources, as well as jitter compensation mechanisms. On the basis of this model, we develop a Holographic DQN-based Resource Scheduling (HDRS) algorithm, which simultaneously optimizes for minimal resource consumption and strict latency guarantees. The simulation results indicate that HDRS outperforms the baseline methods, reducing resource consumption by 23% while maintaining strict latency constraints and achieving a success rate of 95% task deployment under resource-constrained conditions.