For users requesting popular contents from content providers, edge caching can alleviate backhaul pressure and enhance the quality of experience of users. Recently there is also a growing concern about content freshness that is quantified by age of information (AoI). Therefore, AoI-aware online caching algorithms are required, which is challenging because the content popularity is usually unknown in advance and may vary over time. In this paper, we propose an online digital twin (DT) empowered content resale mechanism in AoI-aware edge caching networks. We aim to design an optimal two-timescale caching strategy to maximize the utility of an edge network service provider (ENSP). The formulated optimization problem is non-convex and NP-hard. To tackle this intractable problem, we propose a DT-assisted Online Caching Algorithm (DT-OCA). In specific, we first decompose our formulated problem into a series of subproblems, each handling a cache period. For each cache period, we use a DT-based prediction method to effectively capture future content popularity, and develop an online caching strategy. Competitive ratio analysis and extensive experimental results demonstrate that our algorithm has promising performance, and outperforms other benchmark algorithms. Insightful observations are also found and discussed.

Edge service caching can significantly mitigate latency and reduce communication and computing overhead by downloading and caching service (application) data from clouds. The freshness of cached service data is critical when providing satisfactory services to users, but has been overlooked in existing research efforts. In this paper, we study the online low-latency and fresh service provisioning in mobile edge computing (MEC) networks. Specifically, we jointly optimize the service caching, task offloading, and resource allocation. To solve the formulated joint online long-term optimization problem, we design a Lyapunov-based online framework that decouples the problem at temporal level into a series of per-time-slot subproblems. For each subproblem, we propose an online integrated optimization-deep reinforcement learning (OIODRL) method, which consists of an optimization stage and a learning stage. In the optimization stage of OIODRL, a quadratically constrained quadratic program (QCQP) transformation and a semidefinite relaxation (SDR) method are utilized; in the learning stage of OIODRL, a deep reinforcement learning (DRL) algorithm is applied. Extensive simulations show that the proposed OIODRL method achieves a near-optimal solution and outperforms benchmark methods.