We study edge computing networks with heterogeneous processors (CPUs, GPUs, FPGAs) integrated into selected routers to enable in-network computing via IPv6-SRv6. This supports application-defined compute-and-forward operations, especially for video streaming tasks like replication, encoding, and transcoding at the edge. To manage limited compute and bandwidth resources, we model the system as an online social welfare maximization problem. Our solution includes: (1) a compact exponential algorithm transforming algebraic constraints into geometric ones; (2) a primal-dual approach using multiround auctions; and (3) a dual oracle leveraging network coding for multicast optimization. Simulations show a 61.2% improvement in social welfare over benchmarks, validating efficient, infrastructure-compatible resource coordination under dynamic workloads.

AIGC-as-a-Service (AaaS) enables diverse and high-quality content creation. Due to the computational intensity and high costs of model inference, efficiently scheduling generation requests is non-trivial for AIGC Service Providers (ASPs). A judicious balance is required between generation quality, limited resources, and service delay, while coping with online arrivals and operational constraints. Existing scheduling systems often neglect the quality metrics of generated content, and fail to deploy the latest architecture in model inference. To address these challenges, we introduce a distributed diffusion framework that reduces resource consumption by sharing inference steps across different inference tasks. On this basis, we establish a quality model using the refined CLIP Score, and formulate the scheduling problem as a mixed-integer nonlinear program. We first develop a prompt similarity-based algorithm to determine the number of shared inference steps within each request group. Adopting a primal-dual framework, we then design an online algorithm to dynamically manage request admission and scheduling, maximizing social welfare of the AIGC ecosystem while ensuring generation quality. Extensive real-world trace-driven experiments demonstrate that our approach improves social welfare by up to 38.2% compared to the state-of-the-art method.

We study resource allocation and task scheduling for mobile artificial intelligence generated content (AIGC) in a three-layer cloud-edge-device network. Escalating industry demand for computational resources presents significant challenges in resource allocation and optimization, particularly for edge-side AIGC, which faces high computational costs and requires advanced techniques for efficient model deployment on mobile devices. Optimal resource allocation in mobile AIGC networks is naturally formulated into a 0-1 ILP, which is proven NP-hard. We reformulate the problem into both its Comp-Exp and dual forms. Then, we design an online auction framework online AIGC task scheduling (OATS) to optimize decisions on instances and time schedules, maximizing social welfare for the AIGC ecosystem. Our analysis demonstrates that OATS achieves high social welfare through appropriate bid acceptance and resource allocation. Simulation results corroborate the theoretical analysis, showcasing the efficacy of our online algorithms.

This work studies video transcoding and multicast in an edge computing network (ECN), where routers are furnished with computing resources. For example, the recent IPv6-SRv6 paradigm enables programmable networking that works in concert with such hardware innovations, provisioning application-defined route selection and realizing compute&forward functionalities. A prominent class of user applications in ECN is video streaming. Edge routers may replicate and transcode video streams while delivering them toward end users for customized services, up to node processing and link transmission capacities. We design an auction-based online optimization framework DOP, which comprises of three components: 1) video access management; 2) dynamic price function design; and 3) transcoding multicast tree (TMT) optimization. We formulate the online social welfare maximization problem, and design an efficient primal-dual framework that simultaneously makes video transcoding, multicast routing, and resource pricing decisions. Through rigorous theoretical analysis, we prove DOP guarantees truthfulness and achieves a good competition ratio. Extensive simulations further verify that social welfare increases by about 12.9%, in comparison to benchmark algorithms.