Search

Consumer digital ecosystems include a large volume of different types of applications, and those applications are usually deployed in industrial cloud computing systems. Currently, microservices are one of the most prevailing architectures for industrial clouds. Similar to other architectures, microservices may also produce failures, so failure diagnosis for microservices becomes an inevitable problem in industrial clouds. A majority of existing methods focus on statistical analysis for monitoring data or system topological structure. However, because these methods usually only harness service-level or machine-level metrics, they cannot complete fine-grained failure diagnosis, increasing the running risk of microservice-based industrial clouds. To tackle this issue, in this paper, we design a novel graph structure to represent failure dependencies, especially the heterogeneity, and name it as the heterogeneous failure dependence graph (HFDG). We propose a framework to inform engineers which type of and where failures occur in industrial clouds. The HFDG can be used to mine the propagation of failures between different types of components. We also propose a novel neural network model based on attention mechanism and heterogeneous graph neural network, to fully leverage the metric data and HFDG. We performed experiments on three large-scale public datasets from real-world microservices-based systems. The experimental results demonstrate the superior performance of our model compared to well-known baselines.

In the wake of developments in consumer electronics, electric vehicles are gradually becoming representative carriers of next-generation consumer electronics. Electric vehicles incorporate 5G and artificial intelligence (AI) technologies, which significantly enhance the driving experience. Autonomous driving, as a widely supported feature of most electric vehicles, allows these vehicles to make decisions automatically, reducing safety hazards and enhancing driving efficiency. There has been a growing interest in cost-effective monocular vision solutions to further improve the viability of this technology. However, the limited imaging capabilities of monocular cameras make it challenging to extract explicit depth information from images, which is crucial for locating objects in a scene. In this paper, based on a location-aware attention mechanism and an importance-aware detection head, we propose MonoLI, a monocular 3D object detection method for precisely locating objects. First, the location-aware attention mechanism can perceive location information in both the spatial and channel dimensions. This helps the network learn a more suitable global feature representation for the detection task. Second, the importance-aware detection head considers the difference between auxiliary and target tasks, as well as the importance of different branches in target tasks. In addition, to enhance the applicability and deployability of the method, we introduce partial convolution blocks, which improve the performance and effectively reduce the parameter size and computational complexity. Finally, experiments conducted on the KITTI benchmark dataset show that our method outperforms many well-known baseline methods in terms of 3D object detection and bird’s-eye view (BEV) evaluation.

Consumer electronics such as mobile phones, wearable devices, and vehicle electronics use many intelligent applications such as voice commands, machine translation, and face recognition. These applications require large inference workloads to perform intelligent tasks, which are often completed using deep neural network (DNN) models. Traditional approaches rely on pure cloud computing, with consumer devices collecting data and cloud computing platforms completing inference tasks. In real life, the workloads of these applications are not fixed and are likely to exhibit fluctuations or unexpected surges, increasing the workload of cloud computing platforms. Simply increasing server resources often leads to resource waste. Therefore, a dynamic resource scheduling method is needed. In this paper, a token bucket-based dynamic batching (TBDB) algorithm that maintains throughput while reducing latency and increasing device utilization, especially for large volumes of requests, is proposed. Our work includes the following achievements: 1) We employ the token bucket algorithm to determine the workload, considering the concurrency and frequency of the data. We dynamically vary the maximum batch size (MBS) that will trigger the inference process for the next batch. 2) A low-coupling mode architecture that can be embedded into various consumer electronics in a plug-and-play manner is designed. 3) The performance of the electronic devices and the maximum latency are studied to provide guidance for setting hyperparameters. Finally, we evaluate the effectiveness of our method in three consumer electronic scenarios and present a theoretical analysis for setting hyperparameters in different scenarios.

The emergence of the Internet of Vehicles (IoV) introduces challenges regarding computation-intensive and time-sensitive related services for data processing and communication. Limited resource availability increases the processing latency and may cause application interruption due to the mobility of vehicles. To address the real-time requirements of users and tasks, mobile edge computing (MEC), in which data are processed at the network edge, has been proposed to collaborate with the cloud to provide better performance. However, the offloading strategies proposed previously have some shortcomings in addressing issues such as task dependency and resource competition. In this article, we propose a novel offloading strategy for MEC, Com-DDPG, in which multiagent reinforcement learning is used to enhance the offloading performance. Within the IoV transmission radius, multiple agents work together to learn the changes in the environment, such as the number of mobile devices and the queue of tasks, and take appropriate action in the form of a strategy for offloading to an edge server. First, we discuss models of task dependency, task priority, and resource consumption from the perspective of server clusters and multiple dependencies among tasks. In the proposed method, the communication behavior among multiple agents is formulated; then, the policy determined through reinforcement learning is executed as an offloading strategy to obtain the corresponding results. Second, to enhance the communication of information among multiple agents, a long short-term memory (LSTM) network is employed as an internal state predictor to provide a more complete environmental state, and a bidirectional recurrent neural network (BRNN) is used to learn and enhance the features obtained from the agents' communication. Finally, experiments carried out based on the Alibaba Cluster Dataset are presented. The results show that our method is superior to baseline methods in terms of energy consumption, load status and latency.