In resource-constrained emergency networks, task scheduling algorithms play a crucial role in allocating computational tasks and managing data transmission. Currently, centralized scheduling algorithms face the risk of single-point failures and incur significant overhead in maintaining global computational node states and network topology. Distributed scheduling algorithms typically assume a fixed network topology, making them less adaptable to highly dynamic environments with constantly changing node capabilities and network conditions. To address these challenges, this paper proposes a Multi-agent Deep Deterministic Policy Gradient (MADDPG)-based distributed task scheduling algorithm (MATS). The algorithm deploys an agent on each computational node, which only perceives real-time states of adjacent nodes and links, thereby reducing the overhead of maintaining network states. This paper designs an asymmetric multi-agent structure to accommodate computational nodes with varying performance, introduces a dual-buffer pool structure to accelerate model convergence, and develops an agent action mechanism independent of node scale to enhance adaptability to dynamic network topology changes. Experimental results demonstrate that MATS significantly outperforms existing centralized and distributed approaches in handling node dynamics while achieving task average processing latency comparable to optimal centralized algorithms.

Vehicle Edge Computing (VEC) has emerged as an efficacious paradigm that supports real-time, computation-intensive vehicular applications. However, due to the highly dynamic nature of computing node topology, existing scheduling algorithms need to more effectively apprehend the characteristics of fine-grained task topologies and network topologies. Moreover, they require significant communication overhead and training costs, making them inadequate for fine-grained task scheduling in vehicular networks. In response, our research explores fine-grained task scheduling issues within VEC scenarios, proposing a scheduling algorithm based on Graph Neural Networks and Federated Learning (FL-GNN). This algorithm maintains a global scheduling model that periodically aggregates local scheduling models deployed on Roadside Units (RSUs) and high-performance vehicles. Furthermore, to enhance the model’s ability to perceive topology and expedite the convergence rate, we incorporate a graph neural network layer in each local model to preprocess the raw state of the VEC environment. Lastly, we construct a simulation platform and implement multiple competitive solutions, demonstrating the superiority of the FL-GNN algorithm in aspects such as reducing the average task delay, balancing the load, and improving the task scheduling success rate.

The efficient automated network management methods for mobile operators (for example, mobile traffic prediction) are important goals for future mobile networks. Nevertheless, accurately predicting mobile traffic across an entire city is a challenge that requires the consideration of both prediction accuracy and computational complexity, especially in terms of the thousands of regions involved in high-density and complex base station deployment scenarios for 5G/beyond 5G/6G networks. To solve this problem, this study proposed a novel deep learning network based on the transformer, i.e., a spatial-temporal downsampling neural network (STD-Net), which can dynamically and simultaneously exploit the temporal, local, and global spatial dependencies of mobile traffic. To reduce the computational complexity in spatial domains and achieve a balance between generalization (for all regions) and fitness (for each region), the model decomposes a city into patches and focuses on simultaneously exploiting the temporal and local spatial dependencies in each patch via spatial-temporal transformers. This study’s downsampling transformer is responsible for exploiting global spatial dependencies by uniformly sampling mobile traffic throughout all the regions of an entire city. Computing spatial correlations among sampled regions reduces the computational complexity even further. The superior prediction accuracy of the proposed STD-Net model over state-of-the-art baselines was confirmed by experiments on real-world mobile traffic datasets. The effectiveness of each module was tested to further validate the rationale and feasibility of the proposed STD-Net. Analyses of the computational complexity of the STD-Net revealed that its computational cost contains the same quadratic complexity as vanilla transformers.

In a cloud distributed system, machine failure or region failure is a very common scenario. Data replication is a key technique for ensuring data availability. However, Objects are usually assumed independently by distributed systems, despite, a user-level task typically requests multiple data objects. This paper studies the effect of data placement on the availability of user-level tasks from a theoretical perspective, and finds the best and the worst placements which can provide the highest and the lowest availability for user-level tasks in a cloud distributed system. This paper also gives a novel algorithm called SPOverlap (S Parts Overlap), which provides a tradeoff between task availability and other system performance.

Scalable 3D video P2P streaming systems can supply diverse 3D experiences for heterogeneous clients with high efficiencies. Data characteristics of the scalable 3D video make the P2P streaming efficiency more depends on the data segmentation algorithm. However, traditional data segmentation algorithm is not very appropriate for scalable 3D video P2P streaming systems. In this paper, we propose a Playback Length Changeable 3D video Segmentation (PLC3DS) algorithm. It considers the particular source-data characteristics of scalable 3D video, and provides different error resilience strengths to video and depth as well as layers with different importance levels in the transmission. The simulation results show that the proposed PLC3DS algorithm can increase the success delivery rates of chunks in more important layers, and further improve the 3D experiences of the client. Moreover, it improves the network utilization ratio remarkably.

The PLCS (Playback Length Changeable Segmentation) algorithm was proposed to improve the success delivery rate of chunks in some important layers including the base layer for SVC based P2P streaming systems. In this paper, we make further research and propose a corresponding chunk scheduling algorithm named PLCCS (Playback Length Changeable Chunk Scheduling). In PLCCS, there are two scheduling windows on clients, including General Window and Emergent Window. Different scheduling strategies are adopted in the two windows to ensure the efficiency of data distribution as well as the playback fluency. Moreover, an adaptive strategy is used in General Window to accommodate the change of the network state. We evaluate the performance of PLCCS through simulations. The results show that compared with two representative EPLS based chunk scheduling algorithms, PLCCS can achieve better video quality on clients by at least 40%, and it decreases the useless packet ratio on clients by at least 15 times. Moreover, the adaptive strategy used in PLCCS makes it suitable to be applied in SVC based P2P streaming systems in heterogeneous network environments.

In this paper, we first probe into the data characteristics of SVC and analyze their impacts on P2P segmentation algorithms. According to our studies, the equal playback length segmentation (EPLS) algorithm, which is widely used in current SVC-based P2P streaming systems, does not consider the data characteristics of SVC adequately and thus is easy to cause significant degradation of video quality on clients. Then we propose a novel segmentation algorithm named PLCS (playback length changeable segmentation) for SVC-based P2P streaming systems. In the PLCS algorithm, the playback length of chunks in each layer is determined according to the level of importance in combination with the size of the layer. We evaluate the performance of the PLCS algorithm through experiments and simulations. The results show that compared with the EPLS algorithm, the PLCS algorithm can increase the transmission success rate of layers with higher level of importance and thus can improve the video quality on clients significantly, especially in networks with high packet loss rate.

We propose a novel segmentation algorithm TSESS and a novel chunk scheduling algorithm TSESCS for SVC based P2P streaming systems. TSESS splits each layer into equal sized chunks and tags each chunk with a time-stamp. TSESCS is based on TSESS and adopts two scheduling windows with adaptive strategies. Compared with traditional segmentation algorithm and chunk scheduling algorithms, our TSESS and TSESCS algorithms can achieve better video quality on clients in heterogeneous network environments.