Research On The Resource Management And Service Collaboration For Information Service Cloud Environment

The rapid development of technologies such as the Internet of Things,communication networks,and military information has not only created a large number of terminal devices,but also led to an explosive growth of battlefield data.How to make full use of data resources and how to quickly extract valuable and shareable information from massive heterogeneous complex data has become the key to victory.The information service cloud environment relies on the shared information technology infrastructure,uses cloud computing or edge computing models to analyze and process big data,and provides different levels of data services to ensure seamless,interoperable,and efficient end-to-end information sharing across the military.Currently,the information service cloud environment for big data applications presents four major characteristics:competitive use of resources,high time-sensitive service requirements,diversified service requirements,and high environmental uncertainty,which brings a series of great challenges.To this end,this dissertation mainly focuses on resource management and service collaboration in the information service cloud environment,and studies the service delivery collaboration,and through the optimal scheduling of distributed resources,ensures that the core functions of the information service cloud environment continue to be effective.The specific research content is as follows:Aiming at the problem of different tenant services competing to use network resources,this dissertation proposes a multi-tenant network resource sharing method to improve the performance of upper-layer services.Specifically,the overall data transmission rate of tenants is measured by introducing progress indicators,which reflect the slowest rate at which tenant application services can complete their data transmission.By maximizing the progress of tenants,performance such as the execution time of upper-layer application services can be optimized.On this basis,we propose a network sharing method based on application awareness.By jointly optimizing the process of subtask placement and bandwidth allocation,we maximize the progress of all tenants and maximize network utilization under the fairness of advantageous resources.rate.For terminal devices with limited computing resources,this dissertation proposes a computing resource optimization method for big data clustering to improve the efficiency of unsupervised clustering detection.Specifically,by introducing the convex clustering method,the clustering problem is formulated into a convex optimization form to improve the robustness of the clustering results;on this basis,the similarity between the object instances is used for clustering operations,which is significant Reduce the computational complexity to adapt to large-scale high-dimensional data;finally,two efficient and improved algorithms are designed to solve the new convex clustering model,namely the alternating direction multiplier method and the nested Mirror Prox algorithm.In view of the limited resources at the edge and the use of resource bidding,this dissertation proposes an adaptive resource management method at the edge to reduce the risk and cost of resource preemption.Specifically,the concept of virtual machine portfolio is proposed.It depends on the service’s risk tolerance and cost sensitivity.The virtual machine portfolio has configurable overhead and availability.Secondly,in order to measure and quantify risks,we learn from modern investment portfolio theory.The"mean-variance model" uses the standard deviation of returns to calculate risk,and defines returns as the cost saved by using these examples;finally,it proves that the proposed resource allocation model belongs to a quadratic convex optimization problem,and proposes a Improved completely non-projection near-end stochastic gradient method to solve.Aiming at the dynamically changing service requests of end users and the complicated and changeable edge environment,this dissertation proposes a reliable and efficient service collaboration method at the edge layer to reduce service response time.Specifically,firstly,the system architecture of mobile edge computing network is introduced,and the research motivation of edge service collaboration under this architecture is given;secondly,a risk aversion-based edge service collaboration service collaboration model and problem formulation are proposed,considering The high computational cost and limited battlefield resources further transform the risk aversion model into finite and component optimization.Finally,an edge service collaboration algorithm based on variance reduction is proposed.The theoretical analysis proves that this algorithm can achieve linear convergence speed and significantly reduce computational complexity Performance,large-scale numerical experiments have also confirmed its effectiveness.In response to the unique requirements of cross-layer deployment and use of services,this dissertation proposes a collaborative distribution mechanism of "edge-cloud"services,which effectively improves the service capabilities of the cross-layer model through joint optimization.Specifically,first introduced the three-tier server cluster architecture of the cloud environment,and analyzed the key factors affecting service collaboration,thereby leading to the joint optimization of delivery paths;secondly,the Markov decision process model was used to capture the dynamic changes of the system,and proposed DeepDelivery adaptive deep reinforcement learning method is used to allocate delivery paths;finally,a group-based shrinking method is further designed to reduce the corresponding state space.Extensive simulation experiments confirm that DeepDelivery significantly reduces the service collaboration delay while increasing system utilization.This dissertation focuses on the four maj or challenges in the construction of the information service cloud environment,mainly centering on resource management and service collaboration,to ensure the continuous and effective operation of the core functions in the information service cloud environment,and to improve the ability to sustainably guarantee in a complex cyber threat environment.