Future wireless networks, such as 6G, are envisioned to deliver ultra-reliable, high-quality services with ultra-low latency and dynamic connectivity across heterogeneous environments, driving the adoption of edge–cloud collaborative architectures. Within this paradigm, container-based microservices, with their lightweight, modular, and portable characteristics, offer an effective foundation for scalable and adaptive service provisioning in heterogeneous wireless networks. The layered architecture of microservices facilitates efficient resource management through layer scheduling and caching. However, dynamic service requests and diverse container layers pose major challenges for layer-aware service provisioning in future wireless environments. These includetime-exceeded offline service provisioning, tangled microservice orchestration, andlayer cache redundancy. To address these challenges, we propose Tri-Ring, an asynchronous online provisioning framework for future wireless networks, empowered by large AI models (LAMs). The framework optimizes request dispatching, orchestration, and layer updates across three timescales. At the small timescale, we formulate request dispatching as a linear programming (LP) subproblem. At the medium timescale, the estimator-assessor algorithm manages microservice orchestration, where a diffusion-enhanced prediction model serves as the estimator to predict layer caching strategies. Moreover, submodular optimization serves as the assessor to determine deployment and scheduling. At the large timescale, we introduce the age of layer (AoL) to guide the pruning of infrequently accessed cached layers to reduce storage overhead. Comprehensive evaluations on real-world datasets demonstrates that Tri-Ring outperforms existing baselines, increasing utility by 44.78%, reducing microservice startup time by 78.64%, and optimizing storage resources by 36.38%.

Large language models (LLMs) have emerged as a transformative advancement in artificial intelligence (AI). To fully leverage their potential, Docker containers, serving as a lightweight, portable, and isolated framework, facilitate the seamless deployment of LLM-based applications. However, the deployment of containerized AI microservices faces challenges such as heavy network loads, delayed image loading, and redundancy. In this paper, we introduce MetaPipe, an innovative incremental deployment approach for containerized AI microservices in edge cloud environments. MetaPipe aims to optimize startup times through a dynamic workflow that incorporates proactive layer pre-fetching and reinforcement layer re-scheduling. The proactive pre-fetching reduces service deployment time through layer caching prediction and pre-scheduling before requests arrive, while the reinforcement re-scheduling addresses inaccuracies by dynamically adjusting layer scheduling strategies after requests arrive. Extensive experiments on realworld datasets show that MetaPipe significantly outperforms traditional methods, achieving 83.58% reduction in initialization startup time and 85.58% reduction in cold startup time. These results highlight its effectiveness in enhancing the performance of AI microservices deployment within edge cloud environments.

The rapid growth of distributed service infrastructures has promoted container-based deployment as a lightweight and flexible approach for large-scale service delivery. To enhance the trustworthiness of service deployment, blockchain has been incorporated into container networks as a decentralized trust layer. However, most existing blockchain solutions still rely on coarse-grained trust models that neither capture the evolution of node or layer credibility nor incorporate the layered structure of container images into deployment decisions, which may lead to services being deployed on low-reputation nodes and to the propagation of untrusted layers across the network. To address these limitations, this paper proposes ReTainer, a reputation-aware containerized service deployment framework in blockchain networks. We formulate the deployment problem as a joint optimization and decompose it into a linear-programming request routing subproblem and a service orchestration subproblem, which simultaneously covers service deployment, layer precaching, and layer re-scheduling. A hierarchical trust model captures the temporal evolution of node-level and layer-level credibility, and the resulting reputation priorities are used to select the top-k trustworthy layers for pre-caching. We then formulate the service activation and layer re-scheduling as a submodular optimization problem over a p-extendible system. Experimental evaluations demonstrate that ReTainer improves utility by up to 13.56% and reduces startup latency by 43.31% compared with existing deployment schemes.

Libo Qin, Qiguang Chen, Tianbao Xie, Qixin Li, Jian-Guang Lou, Wanxiang Che, Min-Yen Kan. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.