This work presents a high-efficiency, high-density, and high-speed on-substrate-integrated buck converter for power delivery in mobile devices. This converter employs an advanced on-substrate packaging technique, which integrates all the power devices and capacitors within a compact$5\times 6\times 0.3$mm footprint, making it particularly suitable for space-constrained mobile applications that require fine-grained power management for digital cores. The proposed converter features a two-stage, two-phase architecture and cooperative dynamic frequency control, achieving a 19.1% efficiency improvement under light-load conditions. Fabricated in a 65-nm CMOS process, the converter operates at an 8-MHz per-phase switching frequency and delivers an output voltage range of 0.5–1 V. This design achieves a peak efficiency of 91% and maintains efficiency above 80% over a load range of 0.2–7 A. The power density reaches 0.93 W/mm2.

RDMA is a kernel-bypass and transport-offload technology that provides high throughput and low delay for datacenter networks, and DCQCN is the default and most widely used congestion control algorithm in large-scale RDMA networks. DCQCN involves over 10 parameters at RNICs and switches, and their settings significantly affect network performance, currently relying heavily on exhaustive manual tuning. Although some automatic methods are proposed to tune a subset of DCQCN parameters, none of them comprehensively address all parameters at both RNICs and switches, resulting in compromised network performance. In this paper, we propose Paraleon, an automatic and adaptive system to tune DCQCN parameters comprehensively. We design a millisecond-level sketch-based monitoring mechanism for accurate network-wide measurement, which collects runtime metrics as feedback to guide the tuning process. We also analyze the complicated parameter impacts on network performance, and leverage an improved heuristic searching algorithm for timely performance optimization with better efficiency and convergence. We implement Paraleon and conduct extensive experiments in both NS3 simulations and a real-world testbed. The results show that Paraleon achieves$3.8 \% \sim 61.4 \%$higher performance than existing tuning schemes.

Edge Computing is promising for latency-sensitive applications. However, current edge resource scheduling is inefficient. Deep Learning as a Service (DLaaS) provides deep learning methods to optimize the resource scheduling problem, but faces great challenges of security and reliability. On one hand, DLaaS training agents and raw data are exposed to various adversarial attacks. On the other hand, dishonest DLaaS trainers can generate poisoned models to attack the DLaaS system. In this article, we proposeSAPE, aSecure and decentralized DLAaSPlatform inEdge computing. SAPE allows users to submit their tasks, which will be scheduled to the appropriate edge clusters to minimize the task execution time. We formulate the resource scheduling problem and develop the federated deep reinforcement learning (DRL) method to optimize the problem and resist the adversarial attacks of DLaaS. We utilize blockchain and propose a consortium-based verification scheme to improve the reliability of the federated training process, protecting the DLaaS models from being poisoned and compromised. We conduct experiments to evaluate the latency and security performance of SAPE and the federated DRL scheduling policy. The results show that SAPE outperforms the traditional schemes when defending against adversarial attacks towards the DLaaS platform in edge computing.

Datacenter Quantized Congestion Notification (DCQCN) [12] is the default congestion control algorithm for Mellanox RDMA (Remote Direct Memory Access) NICs [2] in RoCEv2 (RDMA over Converged Ethernet v2) networks, one of the most widely used NICs in leading industry companies [4, 5, 7, 9]. In DCQCN, firstly switches mark packets with ECN (Explicit Congestion Notification) when the queue length exceeds ECN thresholds, then receivers respond to ECN-marked packets with CNPs (Congestion Notification Packets), and finally senders reduce transmission rate when receiving CNPs. DCQCN has 10+ parameters at both NICs and switches, including Alpha Update, Rate Increase & Decrease, Notification Point and ECN thresholds [3], and these parameters have a non-negligible impact on the network performance. Our experiments also verify the network performance of common AI (Artificial Intelligence) training workloads in RoCEv2 networks (e.g., all-to-all collective communication) is greatly influenced by different DCQCN parameter settings (§3). Therefore, when deploying applications in practice, the DCQCN parameters need to be carefully tested and tuned to improve the network performance.

Edge computing architectures can help us quickly process the data collected by Internet of Things (IoT) and caching files to edge nodes can speed up the response speed of IoT devices requesting files. Blockchain architectures can help us ensure the security of data transmitted by IoT. Therefore, we have proposed a system that combines IoT devices, edge nodes, remote cloud, and blockchain. In the system, we designed a new algorithm in which blockchain-assisted compressed algorithm of federated learning is applied for content caching, called CREAT to predict cached files. In the CREAT algorithm, each edge node uses local data to train a model and then uses the model to learn the features of users and files, so as to predict popular files to improve the cache hit rate. In order to ensure the security of edge nodes’ data, we use federated learning (FL) to enable multiple edge nodes to cooperate in training without sharing data. In addition, for the purpose of reducing communication load in FL, we will compress gradients uploaded by edge nodes to reduce the time required for communication. What is more, in order to ensure the security of the data transmitted in the CREAT algorithm, we have incorporated blockchain technology in the algorithm. We design four smart contracts for decentralized entities to record and verify the transactions to ensure the security of data. We used MovieLens data sets for experiments and we can see that CREAT greatly improves the cache hit rate and reduces the time required to upload data.

Edge computing is promising to solve the latency issue in the Internet of Vehicles (IoV). However, due to decentralization, traditional edge computing suffers in management, deployment, and security. Containerization relaxes resource deployment and migration problems, but current container scheduling policies are inefficient to process complicated tasks based on directed acyclic graph or DAG structures. In this article, we design a containerized edge computing platform CUTE, which provides low-latency computation services for the Internet of Vehicles. The centralized controller is empowered with resource management and orchestration, and containers are scheduled to appropriate edge servers to optimize the computation delay. CUTE is also integrated with blockchain to improve network security. We formulate the vehicle task offloading and container scheduling problems and develop a heuristic container scheduling algorithm for DAG-based computation tasks submitted by vehicles remotely. We implement and deploy CUTE into the China Mobile Network, and conduct comprehensive experiments and a case study. The experiment results show that CUTE can provide low-latency computation services for vehicular applications and that the heuristic algorithm outperforms traditional container scheduling policies.

Edge computing, especially multiaccess edge computing, is seen as a promising technology to improve the Quality of user Experience (QoE) of many artificial intelligence (AI) applications in the evolution toward Internet-of-Things (IoT) infrastructure. However, the management and deployment of massive edge data centers bring new challenges for the current network. In this article, we propose a new edge-based IoT platform for AI (EBI-PAI), based on software-defined network (SDN) and serverless technology. EBI-PAI provides a unified service calling interface and schedules the resources automatically to satisfy the QoE requirements of users. To optimize performances during incremental deployment, we formulate the deployment problem, prove its complexity, and design heuristic algorithms to solve it. We implement EBI-PAI based on an opensource serverless project and deploy it in real networks. To evaluate EBI-PAI, we conduct comprehensive simulations based on the generated and real-world network topology, and real-world base station data set. The simulation results show that EBI-PAI can greatly improve QoE with the same budget and save the budget to achieve similar QoE. We finally carry out a case study with real user demands, and it further validates the simulation results.

With the development of Internet of Things (IoT), edge computing becomes more and more prevalent currently. However, edge computing needs to deploy a large number of edge servers to reduce the communication latency, which will bring additional costs to the system. Although there exist some idle computing resources at the edge, the owners distrust each other and lack the incentives to contribute to the system. In this article, we propose a new edge computing platform decentralized and trusted platform for edge computing (DeTEC), which provides a unified interface to users, resolves the user's requests to the most appropriate edge server through domain name server, and returns the computational results to the IoT user. To build a trustworthy system, DeTEC integrates the blockchain technology with edge computing, such that the contributions of each participant could be accounted and rewarded. We formulate the task allocation problem, taking both node capacity and reward fairness into consideration, and solve it through a heuristic algorithm. Finally, to guarantee the trustworthiness of computational results, we utilize a police patrol model and try to optimize the system overall reward. We implement DeTEC based on an open source project and conduct comprehensive experiments to test its performance. The results show that our DeTEC system works well in the IoT scenario.

Industrial Internet of Things (IIoT) has been widely used in many fields. Meanwhile, blockchain is considered promising to address the issues of the IIoT. However, the current blockchains have a limited throughput. In this article, we devise an efficient and secure blockchain protocol compacted directed acyclic graph (CoDAG) based on a compacted directed acyclic graph, where blocks are organized in levels and width. New-generated blocks in the CoDAG will be placed appropriately and point to those in the previous level, making it a well-connected channel. Transactions in the network will be confirmed in a deterministic period, and the CoDAG keeps a simple data structure at the same time. We also illustrate the attack strategies by adversary, and it is proved that our protocols are resistant to these attacks. Furthermore, we design a CoDAG-based IIoT architecture to improve the efficiency of the IIoT system. Experimental results show that the CoDAG achieves 164× Bitcoin's throughput and 77× Ethererum's throughput.

Security is one of the critical issues in traditional networks. Software-Defined Networking (SDN) improves the security aspect by separating the control plane and the data plane of networks. To improve the performance of SDN, researchers have designed many advanced controller prototypes and considered the controller placement problem. However, link failures are critical security issues in networks and greatly impact SDN's security. The controller placement problem for link failures is still challenging today. In this paper, we study the SDN controller placement problem for single-link and multi-link failures, respectively. For single-link failures, we develop a heuristic algorithm to address the controller placement problem. For multi-link failures, we introduce the Monte Carlo Simulation to reduce the computational overhead. We conduct experiments with real network topologies, and the simulation results show that the heuristic algorithm can save significantly more time than the optimal algorithm, while achieving good performance.