Research On The Benefit Optimization Of Access Cloud Resources And Persistency Of Security Cloud Access

Cloud computing is an effective means to store and manage data,improve the utilization of computing resources,and provide various computing and IT services through the Internet.Iaa S,Paa S and Saa S are the three main cloud services,and the service market is becoming more and more mature.To maximize the use of hardware resources and reduce costs,cloud providers have widely adopted server integration methods and shared the same hardware resources among multiple applications running on one.In other words,in an open cloud environment,multiple users compete or use limited cloud resources at the same time.In the existing relatively mature cloud environment,there will still be problems such as idle resources,unreasonable scheduling,security of multi-cloud user access to shared cloud data,and persistency of secure cloud access.Specifically,in a cloud environment,idle servers can increase the maintenance costs of the server owner(or organization),i.e.,reduce the organization's revenue.However,any cloud provider wants to get the maximum benefit,and the opposite is the cloud users.Any cloud user wants to purchase the cloud resources with the highest cost-performance ratio possible.The competition among cloud providers is unreasonable,and the price is too high or too low,which will lead to the loss of cloud users or the decline of quality of service.At the same time,when multiple users and applications share resources on the cloud server,information may be leak through hidden channels related to memory.Encryption can help protect data privacy.However,when memory technology shifts to the use of Non-Volatile Memory(NVM),it is a challenge to support the efficient crash consistency of ORAM systems.The specific main work and innovations of this dissertation are as follows:(1)Distributed idle cloud computing resource optimization based on game theory.The main focus is on the private idle computing resources owned by various organizations willing to form ad hoc cloud provider networks and sell services to cloud users.In this case,organizations(ad hoc cloud providers)can not only meet their own needs but also sell their idle computing resources in the form of ad hoc cloud.Ad hoc cloud providers aim to maximize their benefits by adjusting business costs and sales prices.This dissertation proves that there exists Nash equilibrium in the game.To find the Nash equilibrium of the architecture,this dissertation proposes an iterative proximal algorithm(IPA).By analyzing the convergence of the algorithm IPA,it is found that the algorithm conforms to the theoretical proof and is feasible.Experimental results show that the algorithm IPA can not only quickly converge to a stable state,but also the benefits of ad hoc cloud providers have been increased compared to when there is no algorithm IPA.(2)A hybrid game method of demand and response of multi-cloud providers and multicloud users.To solve the demand and response problem between multiple cloud providers and multiple cloud users,while making resource scheduling better and transaction prices more reasonable,this dissertation designs a new service architecture model.This problem is discussed from the perspective of game theory,but it takes a different approach compared with the existing cloud resource pricing game method.In this dissertation,the competition among cloud providers is modeled as a noncooperative game process,while the competition among cloud users is modeled as an evolutionary game process.Finally,the two-game processes are integrated into a cloud environment system to form a hybrid game model.In this hybrid game model,the strategy of each cloud user and each cloud provider is not only affected by the other side,but also by the customer or cloud provider other than themselves.Both sides try to maximize their benefits under each other's influence.In this dissertation,it is proved that the two-game processes can converge to their respective equilibrium solutions(Nash equilibrium and evolutionary equilibrium)when the strategy is proposed.Finally,the strategies of cloud users and providers can be implemented through a new iterative proximal algorithm(NIPA)and distributed iterative algorithms(DIA).Experimental results verify the feasibility of this method and show that this method can benefit multiple cloud providers and multiple cloud users.(3)PS-ORAM: Efficient Hardware-Assisted Persistent Oblivious Random Access Machine(i.e.,ORAM)System for Non-volatile Memory.Cloud computing has developed into a promising computing paradigm.However,protecting application privacy,especially memory access patterns on cloud servers,is still a difficult task.ORAM protocol can realize the privacy protection of access patten.However,when storage technologies shift to non-volatile memory(NVM),new challenges arise in how to effectively support ORAM crash consistency.Research has found that using traditional software-based support increases performance overhead and leads to other information leakage.Therefore,this dissertation proposes PS-ORAM to solve this problem,which studies how to use the storage system based on NVM to persist ORAM data.This dissertation analyzes the design requirements for persistent ORAM systems and discusses the need to maintain crash consistency and atomicity of ORAM data and metadata.Next,we discuss some of the challenges in persistence ORAM system design and propose some solutions to these challenges.Then,an improved on-chip ORAM controller architecture is proposed in this dissertation.Based on the improved hardware architecture of the onchip ORAM controller,this dissertation proposes different design persistence protocols from the software perspective to ensure that the ORAM system can maintain data con-sistency when the system crashes.The proposed PS-ORAM system does not disclose more privacy information when it is implemented.Finally,the feasibility and performance of the PS-ORAM system are verified by experiments.The experimental results show that PS-ORAM can minimize system performance loss when the main memory is NVM.(4)HM-ORAM: A Lightweight Crash-consistent ORAM Framework on Hybrid Memory System.Innovation(3)applies to systems where the main memory is pure NVM.The experimental verification shows that the PS-ORAM system can make the performance loss of the system as low as possible on the basis of the original nonpersistent ORAM system.However,due to current NVM device material and process limitations,the read/write delay is still very large,so the system performance loss will be relatively large.Therefore,the innovation of this work is not only to solve the crash consistency problem of the traditional ORAM system but also to improve the performance problem of supporting crash consistency ORAM as much as possible.In view of the above problems and analysis,this innovation is to use the hybrid memory(DRAM+NVM)architecture to study the lightweight crash consistency support for ORAM.This dissertation introduces the HM-ORAM framework,which makes full use of high-performance DRAM and persistence NVM technology.Analyze the design requirements of the crashconsistent ORAM system,and propose a novel ORAM partitioning scheme,which can reduce the overhead caused by excessive write operations on NVM to achieve data persistence.Without compromising security,the performance of HM-ORAM in nonrecursive and recursive implementations is much better than that of ORAM based on NVM.In addition,HM-ORAM also significantly reduces the write traffic to NVM,thereby reducing the impact on the lifetime of NVM.