To address the problems of low resource utilization and deployment efficiency caused by sudden changes in network states due to large-scale complex network access and the random arrival of service requests during dynamic service function chain (SFC) deployment, a novel SFC deployment method for IoT networks based on large language model (LLM) policy distillation is proposed. First, an SFC deployment framework based on teacher-student agent policy distillation using LLM is constructed. Through the policy distillation mechanism, the MARL-based student agents are guided to efficiently learn the SFC deployment policies generated by the LLM-driven teacher agents. Second, we design a teacher agent composed of three modules: a state perceiver, a task-planning decision-maker, and a result evaluator. The state perceiver predicts node resource availability using an LLM-based spatiotemporal forecasting method. The decision-maker leverages LLM reasoning to generate candidate deployment policies. The evaluator selects policies based on load-balancing metrics. Finally, the student agents, considering constraints such as network resources and latency, build an optimization model aimed at maximizing resource utilization and SFC deployment rewards, and introduce a Teacher-Student Policy Distillation-based Multi-Agent Soft Actor-Critic (TSPD-MASAC) algorithm to solve this optimization problem. Simulation results demonstrate that, in complex network environments with dynamic resource states and diverse service requests, the proposed method achieves more accurate resource state perception, accelerates algorithm convergence, and significantly improves resource utilization and overall deployment performance compared with baseline schemes.

To address the issue of inaccurate synchronization between physical devices and their corresponding digital twins (DT) in Industrial Internet of Things (IIoT), which is caused by sensing errors, unreliable wireless transmission, and outdated information, we propose a DT information synchronization strategy for IIoT based on dual time scale network slicing (NS) orchestration. Firstly, a DT-driven IIoT slicing architecture is proposed to provide isolation for heterogeneous Quality of Service (QoS) requirements. On this basis, to quantify synchronization performance, a DT fidelity model is established, incorporating sensing accuracy, data transmission reliability, and the Age of Information (AoI). To fully utilize the network resources and improve the accuracy of DT synchronization information, a dual time-scale model is constructed, where the large time scale handles DT association and inter-slice resource allocation according to the users’ demand, while the small time scale is responsible for intra-slice scheduling of power, bandwidth, and observation frequency. To solve the formulated optimization problem, we design a hierarchical deep reinforcement learning framework that adopts Deep Recurrent Q-Network (DRQN) and Counterfactual Multi-Agent Prioritized Experience Replay Compound-Action Actor-Critic (COMA-PER-CA2C) algorithms. Simulation results demonstrate that the proposed method significantly improves DT fidelity and resource utilization in various IIoT scenarios.

Combining semi-grant-free (SGF) transmission with non-orthogonal multiple access transmission (NOMA) allows grant-free (GF) UEs to access the network through resource blocks that are solely allocated to grant-based (GB) UEs. A channel gain constraint for GF UEs is introduced to reduce the interference impact to GB UE, which causes an access fairness problems for GF UEs. Moreover, the power collision (PC) interference significantly degrades the interference cancellation performance. To address these issues, this paper proposes a layer SGF-NOMA random access (RA) procedure and a joint power back-off (PB) scheme. In the proposed RA procedure, GF UEs can freely share the GB UE’s resource blocks without channel gain constraints. To suppress the PC interference, the proposed PB scheme optimizes the power distribution in the NOMA group according to the variable interference experienced by the NOMA UEs. To maintain the signal-to-interference balance, more power is transferred to the UEs that suffer more interference. The closed-form expressions of the outage probability, successful access probability and sum data rate are derived via order statistic theory. Both the analytical and simulation results demonstrate that the layer SGF-NOMA procedure and joint PB scheme achieves significant performance gains over traditional schemes.

Rapid signal fluctuations due to blockage effects cause severe risks of handover failures (HOF) and ping pongs (PPs). By reconfiguring line-of-sight (LoS) Links through passive reflections, intelligent reflecting surface (IRS) has the potential to maintain link stability and resolve this issue. However, existing handover (HO) process analyses have not introduced both blockage effects and IRS reflections, thus fail to explore the potential of IRS in this aspect.This paper analyzes the IRS-aided HO process by tracking the Line-of-Sight (LoS) state of moving users, where effects of LoS state transitions are characterized by modifying the state transition probability of HO process. Specifically, LoS states involve LoS, non-LoS, and IRS-reconfigured LoS, and the state transition probabilities are obtained through exact blockage modeling and IRS reflection analysis, taking into account the correlation of adjacent moments. In addition, Markov Chain-based analytical models are designed for the process of HO, HOF, and PP considering HO parameters (Time-to-Trigger and HO margin), whose transition probabilities are integrated with all LoS states. The results indicate that under severe blocking effects, the trade-off of HO parameters becomes ineffective, with HOF and PP probabilities remaining high. However, after introducing IRS, a viable range of HO parameters emerges.

Reconfigurable intelligent surface (RIS) is a promising technology for 6G wireless communication. To learn from former successful deployed technique, a mature and widely deployed solution in 5G networks, the repeater, is a subject worth comparing. This paper presents a comparative study of RIS and commercial repeater through field trials in a typical indoor parking lot scenario. By providing empirical insights rather than algorithmic contributions, the study focuses on assessing their performance in terms of coverage, signal quality, and deployment characteristics. Field trials were conducted to measure key metrics such as reference signal received power (RSRP) and signal-to-interference-plus-noise ratio (SINR) under both point and area coverage scenarios. The results reveal that RIS achieves localized coverage improvements with enhanced signal reflection, whereas repeater excels in wide-area coverage through cascaded digital remote units (DRUs). The findings highlight the distinct advantages and limitations of each technology, providing practical insights for their deployment in indoor environments.

With the rapid development of the Internet of Things (IoT), many IoT devices are accessing the network. However, existing networks cannot fully meet the strict and diverse requirements for delay and reliability in delay-sensitive services. Dynamic changes in service requests and the states of service nodes cause a lack of guaranteed end-to-end (E2E) network slicing delay determinism. To address this issue, we propose a digital twin (DT)-assisted network slicing resource allocation scheme. By integrating DT and network slicing, we first construct a DT-assisted E2E network architecture, and construct the base and mapping models in the proposed architecture. Second, we use the stochastic network calculus (SNC) theory to analyze the E2E delay violation probability and characterize the relationship between delay and service reliability under given traffic arrival distributions and delay constraints. Then, we construct a joint resource allocation problem of time-frequency, computation, storage, and bandwidth resources to maximize the utility of the infrastructure provider while guaranteeing the deterministic delay. Furthermore, a multiagent deep reinforcement learning algorithm in a distributed architecture is used to solve the complex optimization problem, achieving efficient network resource allocation. Simulation results demonstrate that the proposed resource allocation scheme meets the requirements for deterministic delay and enhances system utility.

To address the low utility of resource management strategy due to diversified user Quality of Service (QoS) requirements and inaccurate information synchronization in Digital Twin Networks (DTN), we propose a dynamic slice resource management and information synchronization strategy in Internet of Vehicles (IoV) based on Digital Twin (DT). First, to realize the dynamic resource allocation between slices and users respectively to adapt to network dynamics, we propose a two-level dynamic resource management strategy based on the DT-assisted slicing architecture of IoV. Second, to guarantee the timeliness of user information transmission and realize the accuracy of the resource management strategy, we propose a DT satisfaction evaluation model including DT mapping granularity, DT timeliness, and resource residual rate to quantify the users’ satisfaction with their DTs association. Finally, we establish a joint optimization model for resource management and DT association to maximize system utility. To address the coupling between strategies, we split the optimization problem into service utility and information synchronization utility subproblems. In the service utility subproblem, we propose a Counterfactual Multi-Agents Twin-Actors Soft Actor-Critic (COMATASAC) algorithm, which can perform slice-level and user-level resource allocation and scheduling actions separately. In the information synchronization utility subproblem, we utilize the Branching Dueling Q-network (BDQ) algorithm to solve the dimensionality explosion problem, and implement the association policy between users’ DTs and servers. Simulation results show that the proposed scheme can effectively reduce the latency and improve the satisfaction of DTs deployment while guaranteeing the QoS.

Carrier-sense multiple access with collision avoidance (CSMA/CA) is one of the fundamental medium access control (MAC) protocols for ad hoc networks. As a network increases in size, its throughput degrades substantially due to packet collision. To reduce the collision probability, a nonorthogonal multiple-access (NOMA)-based CSMA/CA (NCSMA) protocol is proposed. The proposed NCSMA protocol employs Zadoff-Chu (ZC) sequences and a NOMA power coefficient allocation (N-PCA) frame to select the NOMA node and schedule the power level, preventing packet collisions caused by random selection. Moreover, the normalized saturation throughput and the average packet delay of the proposed approach are theoretically analyzed, and closed-form expressions are derived. Simulation results show that the proposed protocol improves the normalized saturation throughput by 66.7% and reduces the average packet delay by 35.97% compared to those of the current best CSMA/CA protocol.

Network slicing (NS) enables flexible allocation of the Internet of Things (IoT) network resources through software-defined networking (SDN) and network function virtualization (NFV) technologies. However, dynamic variations in network traffic and resource demands can lead to node overload and failures, necessitating timely virtual network function (VNF) migration to safeguard IoT business Quality of Service (QoS). Addressing the issue of deteriorating QoS due to untimely VNF migration, we propose a digital-twin (DT)-assisted VNF migration strategy based on resource demand prediction. First, a prediction model combining convolutional neural networks, gated recurrent units, and attention mechanisms is proposed to forecast VNF resource requirements. Second, the granularity of DT synchronization information is adjusted to resolve issues of delay and high cost during DT construction. Then, a VNF migration model is constructed to maximize DT utility while reducing network costs and average resource variance. Finally, a multiagent algorithm combining long short-term memory (LSTM) and double deep Q-network (DDQN) is proposed to perform VNF migration based on their priority-driven predicted future resource demands, and a multiagent algorithm based on dueling DDQN (D3QN) and deep deterministic policy gradient (DDPG) is proposed to address the DT association problem with a mixed action space. The simulation results demonstrate that the proposed algorithm can reduce the synchronization latency of the DT, the violation rate of service-level agreements, and the service outage time rate, while improving the network load balancing capability.

Network function virtualization (NFV) and software-defined networking (SDN) provide high-quality services to users of the Internet of Things (IoT). However, dynamic changes in network traffic and service function chain (SFC) resource requirements may result in virtual network function (VNF) migration issue and real-time triggering of VNF migration can cause service delay issues in SDN/NVF-enabled IoT networks. In this article, we propose a digital twin (DT)-assisted VNF migration strategy to effectively address this issue. The digital twin of VNF (DT-VNF) is integrated with a multitask DT migration model based on bidirectional-gated recurrent units (DTMBi-GRUs) to achieve accurate resource demands prediction. Based on this, the VNF migration strategy is formulated in advance to avoid network performance degradation. We focus on the post-migration effects on services, networks, and DTs, so migration plans for DT-VNF and a reassociation scheme are formulated to enable real-time monitoring of post-migration VNF by DT-VNF. Then, an optimization problem is formulated to minimize average network energy consumption, network resource differences, and SDN synchronization delay in order to obtain optimal strategies. In addition, considering the problem’s complexity, it is decoupled into the VNF migration problem and the DT association and migration problem. The collaborative solution involves employing the multiagent proximal policy optimization (MAPPO) and asynchronous advantage actor–critic (A3C). Simulation results confirm the superiority of the proposed algorithms over baseline algorithms.

As an important application scenario for 5G, the Internet of Vehicles (IoVs) achieves extensive and stable connections and real-time information interaction and sharing between vehicles and traffic infrastructure. To address the challenges of low-latency synchronization, high computational overhead, and multidimensional resource scheduling faced by the vehicle digital twin (VDT) construction process within IoVs, we propose a construction scheme for variable granularity VDT. First, a variable granularity VDT construction framework for IoVs is proposed to reduce the communication pressure, edge load, and energy consumption in the network. Second, the VDT utility is quantified from four dimensions: 1) completeness; 2) accuracy; 3) timeliness; and 4) energy efficiency. Then, an optimization model is established with the goal of maximizing the average utility of the system VDT, which involves joint optimization of vehicle-edge association, VDT granularity setting and resource allocation. Due to the complexity of the optimization problem, it is decomposed into subproblems of vehicle-edge association, VDT granularity setting and resource allocation, and solved using matching theory and multiagent deep reinforcement learning, respectively. Finally, simulation results verify that the proposed scheme can effectively improve the utility of VDT in different simulation scenarios while reducing system resource consumption.

Digital twins (DTs), as an effective technology for remote monitoring and management of devices, enhances the intelligence of the industrial Internet of Things (IIoT). Nonetheless, the unreliable and delayed transmission of sensory data in wireless access networks hinders the accurate reflection of DTs on the physical world. In this article, we present an intelligent dual time-scale network slicing strategy utilizing the long-term and short-term trends of network, aiming to make fuller use of network resources and improve the synchronization information accuracy of DTs. Specifically, within the dual time scale slicing framework, this strategy collaboratively optimize slice scaling and sensory information synchronization for DTs, aiming to maximize sensory information satisfaction and minimize the cost of slice reconfiguration and synchronization. First, at large time scales, we utilize slices to provide isolation and address deployment issues for DTs with different Quality of Service (QoS) requirements. At small time scales, we aim to enhance the adaptability of estimation tasks to dynamic environments through more flexible wireless resource allocation, further improving communication performance, and establishing DTs that closely resemble physical entities. Furthermore, to solve optimization problems at different time scales, we propose a two-layer deep reinforcement learning (DRL) framework to achieve efficient network resource interactions, in which the lower-layer control algorithms utilize the prioritized experience replay (PER) mechanism to accelerate the convergence speed. Finally, simulation results validate the effectiveness of the proposed strategy.

Network function virtualization (NFV) and software defined network (SDN) technologies enable flexible traffic scheduling and improve the efficiency of physical resources. However, for latency-sensitive Industrial Internet of Things(IIoT) services in Industry 4.0 and beyond, the inability of the SDN controller to synchronize the resource demand information of virtual network functions (VNFs) in a timely manner can lead to delays in VNF mapping and scheduling strategies. To address this issue, we propose a digital twin-assisted VNF mapping and scheduling algorithm that combines digital twin to assist the SDN controller in collecting data of VNFs. Firstly, we designed a digital twin-assisted and SDN/NFV-based network slicing architecture. Secondly, to reduce the synchronization delays of digital twins of VNFs, we propose a twin service node reassociation mechanism. Next, a digital twin-assisted VNF mapping and scheduling model is constructed under constraints such as CPU, storage, bandwidth resources, and quality of service (QoS) to maximize the service provider’s profit. Finally, we propose digital twin-assisted VNF mapping and scheduling algorithms based on greedy and tabu search to solve the problem. Leveraging digital twins, the SDN controller can obtain accurate resource demand information of VNFs, thereby solving the delay issue in VNF mapping and scheduling strategies. Simulation results indicate that the proposed algorithms yield favorable outcomes in terms of total profit, network service acceptance rate, average system delay of digital twins, and QoS satisfaction.

With the rapid development of cloud computing, IaaS (Infrastructure as a Service) becomes more and more popular. IaaS customers may not clearly know the actual performance of each cloud platform. Moreover, there are no unified standards in performance evaluation of IaaS VMs (virtual machine). The underlying virtualization technology of IaaS cloud is transparent to customers. In this paper, we will design an automated performance benchmarking platform which can automatically install, configure and execute each benchmarking tool with a configuration center. This platform can easily visualize multidimensional benchmarking parameters data of each IaaS cloud platform. We also rented four IaaS VMs from AliCloud-Beijing, AliCloud-Qingdao, UCloud and Huawei to validate our benchmarking system. Performance comparisons of multiple parameters between multiple platforms were shown in this paper. However, in practice, customers' applications running on VMs are often complex. Performance of complex applications may not depend on single benchmarking parameter (e.g. CPU, memory, disk I/O etc.). We ran a TPC-C test for example to get overall performance in MySQL application scenario. The effects of different benchmarking parameters differ in this specific scenario.