The emergence of cloud and edge computing has solidly established the Cloud-Edge collaborative architecture as an essential technology, enabling the deployment of large-scale distributed applications. However, despite the benefits conferred by Cloud-Edge collaborative systems, challenges arise from prolonged operation, rapid evolution, and resource constraints. These challenges may accelerate software aging, resulting in performance degradation, functional disruptions, and increased susceptibility to errors. Software rejuvenation acts as a preemptive and proactive maintenance strategy designed to alleviate the consequences of software aging. Although task migration is a commonly used method for software rejuvenation, addressing software aging in complex Cloud-Edge collaborative systems requires dynamic task migration to adapt to environmental changes. Therefore, to develop a task migration-based rejuvenation strategy with lower latency, energy consumption, and software rejuvenation cost, this paper introduces a method for generating software rejuvenation strategies based on Multi-Agent Deep Reinforcement Learning for Cloud-Edge collaboration systems, named GRCEM. Initially, the environment is modeled to define the state space, action space, and training rewards, meticulously designed to minimize rejuvenation costs for the Cloud-Edge collaborative system. Subsequently, a distributed execution algorithm for deep reinforcement learning is devised, treating each node as an autonomous agent responsible for acquiring knowledge and enhancing rejuvenation strategies through continuous interaction with the system environment. Finally, the rejuvenation strategy is formulated, characterized by its optimal rejuvenation cost. The experimental results demonstrate that the GRCEM method accomplishes software rejuvenation with reduced rejuvenation costs compared to the baseline method. Furthermore, the system’s performance transitions from the aging state to the robust state after rejuvenation, thereby significantly enhancing the availability and reliability of the Cloud-Edge collaboration system.

As a platform for creating, deploying, and managing containers, Docker has long been tasked with handling high workloads, making it highly susceptible to aging-related bugs. As these bugs accumulate, the system may exhibit anomalies such as increased resource utilization, task scheduling failures, and response time delays. At this juncture, the system is subject to software aging. If left unresolved, this problem may escalate to more severe consequences such as system crashes and downtime, significantly diminishing the availability and reliability of the system. In order to address the software aging and restore system performance, it has become an urgent problem to accurately determine the aging state of the Docker platform and generate targeted rejuvenation operations reasonably and effectively. Therefore, this paper proposes a synthesis method for determining the aging state and generating rejuvenation operations, named FOQL. Firstly, the FS-OWA algorithm is employed to analyze resource usage according to the varying degrees of aging states, accurately determining whether the system is in an aging state. Secondly, if the system enters an aging state, the Q- Learning algorithm evaluates the value of each rejuvenation operation based on the degree of aging and the cost of rejuvenation operations (such as downtime), ultimately generating the optimal operation. Finally, the experimental results show that, in determining the aging state, the recognition accuracy of the FS-OWA algorithm reached 99.3%, surpassing baseline algorithms by up to 16.52 %. In generating rejuvenation operations, Q-learning algorithm generates a Q-table containing the value of each state-action pair. Based on this table, the optimal rejuvenation operation can be selected for execution. In conclusion, the utilization of the FOQL method effectively mitigates the aging problem and ensures the service quality of the system.

The Cloud-Edge collaborative system, which combines the advantages of cloud computing and edge computing, has become the preferred architecture for large-scale distributed systems. However, prolonged high-load operation of Cloud-Edge collaborative system may result in software aging, significantly impacting the reliability of the Cloud-Edge collaborative system, especially in resource-constrained edge environments. Proactive rejuvenation can help restore system robustness, but it comes with costs and affects normal system operation. Determining the appropriate rejuvenation timing and implementing an efficient rejuvenation strategy are crucial. Since the reliability of the edge system is closely related to the stable operation of the entire Cloud-Edge collaborative system, this paper proposes a comprehensive rejuvenation model named SM-OLR. This model calculates the rejuvenation time for the edge system and performs the rejuvenation operation, which is divided into two stages. The first stage involves determining the rejuvenation timing. A Semi-Markov model is used to represent the system state. Specific distribution functions are fitted based on measured values to accurately describe the system's state transitions. This enables a more scientific modeling of the system's state and ensures precise calculation of rejuvenation timing. The second stage focuses on the rejuvenation strategy. In the Cloud-Edge collaborative environment, interactions between edge and cloud, and edge and edge are easily facilitated, making task offloading a highly effective rejuvenation method. The paper adopts task offloading as the rejuvenation strategy. The reinforcement learning algorithm SARSA is employed to dynamically decide the offloading decision for specific tasks. Experiments were conducted in KubeEdge, a representative Cloud-Edge system. The results demonstrate that SM-OLR incurs lower rejuvenation overhead compared to traditional reboot rejuvenation, and the edge system can continue to provide services during the rejuvenation process. By performing effective rejuvenation operations, the performance of the edge system can be improved by 57 %.

Android systems are prone to software aging due to the accumulation of numerical errors and storage-related bugs during long-term operation, resulting in gradual performance degradation and sudden system hang-ups. Thus, it is very critical to accurately identify the aging state for improving the running reliability of Android systems. In this paper, we propose a novel software aging state identification method, named EWDLL. It first introduces the exponential Weibull distribution to simulate the aging state transfer process of the Android system, then it uses Fuzzy Analytical Hierarchy Process (FAHP) to weight the model parameters and resource utilization parameters. Finally, the weighted dataset is fed into the LightGBM-LR model to identify the software state. The experimental results show that our EWDLL method performs better in identifying the software aging state for Android system, i.e., it is 0.86% to 1.09% higher in identification accuracy than the pure LightGBM-LR model, about 10.00% and 4.54% to 4.95% higher than the traditional models KNN and RF, and 1.97% to 3.09% higher than single LightGBM model. Compared with the LR model, it has a maximum accuracy improvement of about 33.29% to 35.64%.

Because Virtual Machine Monitor (VMM) runs continuously with high load for a long time, the accumulated errors in the system can easily lead to software aging problems such as performance degradation, so that it can not provide high-quality software services. However, software rejuvenation techniques can improve the performance of software systems by cleaning up the aging factor. Therefore, how to accurately predict the occurrence time of software aging in VMM system is a crucial and valuable problem. This paper proposes a novel software aging prediction model, called the ACLM model, which integrates the Attention mechanism, the Convolutional Neural Network (CNN) and the Long Short-term Memory (LSTM) model. It can extract the temporal and spatial features of the time series more quickly and effectively. The experimental results show that compared with other models, the ACLM model has improved by 0.97% to 9.33 % and 1.48 % to 5.49 % on the MAE and the RMSE, which demonstrates that the ACLM model has higher accuracy in predicting software aging of the VMM.

Cloud services typically compose of multiple distributed software components that communicate with each other through web service interfaces in the cloud environments. During their long time running, the accumulation of cloud software internal errors or large consumption of computing resources will very likely lead to software aging problems. In order to solve this problem, software rejuvenation technology is proposed to prevent them from causing more serious failures by restarting the services running. In the research field of software aging and rejuvenation for cloud services, how to accurately predict the cloud resource consumption in the aging software system for determining suitable time to perform rejuvenation is a significant and indispensable issue. In this paper, a novel hybrid aging prediction model named CSSAP is proposed, which well integrates the Autoregressive Integrated Moving Average (ARIMA) model and Long Short Term Memory (LSTM) model for better fitting the linear pattern and mining the nonlinear relationship in the time series of computing resource usage data for cloud services. The experiments results show that through such hybrid and unified time series analysis, our CASSP prediction method has 4% to 71% improvements in MAE evaluation criteria and 6% to 66% improvements in RMSE evaluation criteria under different time series scenarios compared with single model used, that is, the more accurate and more comprehensive aging prediction results achieved by CSSAP is definitely conducive to perform more effective and more efficient software aging and rejuvenation for cloud services.