Key Technologies On Resource Scheduling And Task Offloading In Collaborative Computing With Cloud And Client

In the era of smart connections of everything,new network technologies and smart+ technologies are developing and integrating rapidly.Emerging applications are arising,smart terminals are increasing dramatically,and user data is exploding.In complex and changing network scenes,for the emerging application requests of massive users,cloud computing with centralized data processing suffers from long transmission distance,high response time,easy network congestion,increased task delay,and rising system energy consumption.For edge computing with decentralized data processing,due to the time-space variation of user tasks and multiple resource constraints such as edge devices,the load of edge devices is easily imbalanced,making it difficult to guarantee the quality of service(QoS)and the system service energy efficiency.The current network computing meet serious challenges in providing fast and efficient computing services.To solve the performance problem in current network computing,meet the emerging application requirements,and provide fast and efficient computing services for users and service providers,we studied the key technologies of resource scheduling and task offloading in collaborative computing with cloud and client.First,we studied collaborative resource scheduling in cloud computing.Secondly,we introduced the edge layer in cloud-client to study edge-cloud collaborative computing task offloading.Then,we considered the resources required for task execution to study edge-cloud collaborative resource deployment.Finally,we combined resource deployment to study edge-cloud collaborative task scheduling.For the research content,we proposed innovative theories and solutions in four aspects.1.A multidimensional QoS collaborative resource scheduling with a stakeholder perspective approach is proposed.For large-scale heterogeneous clouds,scheduling resources to execute tasks suffers from low user service quality and high system cost.First,a two-level resource scheduling model combining virtual and physical resources is proposed for task execution.Then,a multidimensional QoS cloud computing architecture with user and service provider stakeholder perspective is designed,and quantifies multidimensional QoS degree combining cloud task execution time,cloud task completion time,cloud system energy consumption,cloud system availability,and cloud system economy.Finally,based on multiple Greedy ideas,a multidimensional QoS collaborative resource scheduling algorithm(MQoS)is proposed by considering minimum L,maximum S,and maximum R.Compared with existing methods,MQoS can optimize the physical and virtual resource scheduling for both user and service provider,and improve the multidimensional QoS degree and reduce the cloud data center load balancing difference in cloud computing scenes with different task differences and varying resource differences.2.A collaborative computing task offloading with swarm intelligence evolution approach is proposed.For delay-sensitive computing task offloading there are problems of long delay,high energy consumption,and unbalanced load.First,considering the task transmission and execution,quantifying the task delay and energy consumption,and constructing an edge-cloud collaborative computing model.Then,adapting to edge-cloud collaborative computing,a multi-strategy improved sparrow search with swarm intelligence evolution algorithm(MISS)is proposed,which optimizes the task offloading by updating the location search of discoverers,followers,and agitators through flier momentum,nonlinear search factor and sin/cos perturbation quantum,and adaptive adjustment of agitator combining location search deviation entropy and nonlinear warning coefficient,respectively.Finally,a heuristic task offloading with MISS algorithm(HTOM)is proposed,which combines different delay constraints of task maximum completion period and delay relaxation variables,and considers the energy consumption of timeout penalty to further optimize the task offloading decision of edge devices.The experimental results verify that HTOM can efficiently offload computing tasks and outperforms existing comparison schemes in terms of total task energy consumption,average task completion delay,and node load balance degree.3.An edge-cloud collaborative resource deployment with task prediction approach is proposed.For the problem of multiple resource constraints and time-space variation of tasks,the task local execution rate of edge servers is low.First,a two-dimensional time series-based task prediction algorithm(TSTP)is proposed,which uses moving average prediction with delayed deviation correction to obtain horizontal time series,plus the mean value of vertical time series with different periods,to classify and aggregate tasks in the cloud.Then,a resource deployment with TSTP prediction algorithm(RDTP)is proposed,which predicts and ranks tasks in descending order of frequency,combined with task delay threshold determination,to classify and optimize the deployment of resources required for delay-sensitive and non-delay-sensitive task operation.Experimental results show that RDTP can effectively predict task deployment resources,and its average user task deviation and average user task hit rate are better than similar methods under different tasks and edge servers.4.A Pareto-optimized collaborative task scheduling combining resource deployment approach is proposed.For solving the problem of low user service quality and system service effect for task scheduling considering resource deployment under complex network changes.First,a task scheduling with Pareto improvement algorithm(TSPI)is proposed,which solves m group task pre-scheduling scheme in two stages using stochastic greedy approximation algorithm for the objectives of user service quality and system service effect,and then optimizes the task scheduling scheme of edge servers by Pareto progression.Then,a Pareto-optimal task scheduling combined with RDTP resource deployment algorithm(RDTP-TSPI)is proposed,which further optimizes edge-cloud collaborative task scheduling based on predicting tasks and optimizing resource deployment of edge servers.The experimental results show that RDTP-TSPI can adapt to complex network changes,and its average task completion time,overall system service effect,and total task delay rate outperforms the comparison schemes for different task Zipf distribution parameters and various network sizes.