Research On Key Problems Of Quality Of Service In Cloud Storage

With the progress of science and technology, the information technology has been de-veloping with each passing day. Especially, the development of the internet has contributed greatly to the change of human life. Since the release of Amazon Web Service (AWS) sys-tem by Amazon from2006and the proposal of the concept of cloud computing by IBM and Google in2007, cloud computing has gradually gone into people’s life, making it pos-sible for people to consume computation resources like they consume water and electricity according to their demand in the form of pay-as-you-go.As we all know, cloud storage is well accepted by people according with the explosive growth of today’s society information. In addition to the early adopters from the internet companies, more and more enterprises, organizations, and individuals begin to rely on cloud storage to hold their huge digital information. Some challenges have been brought out, such as data storage costs, reliability and availability with the scale expansion of cloud service. Meanwhile, the cloud storage service users will take into consideration various factors before migrating data to clouds, for instance the overhead, quality of service, data availability, especially the follow-up cost of continuous service. In order to make the right decision, whether enterprises or individuals will give a deep thinking before involving in cloud services.To address the problems with the bottleneck of public cloud storage performance and the lack of accuracy of the budget for both sides between cloud storage providers and users, this paper is to deal with the key problems of quality of service in cloud storage including public cloud storage performance optimization, cloud storage system reliability and I/O resource scheduling strategies. In addition, a big challenge of IaaS providers is how to account the cost and charge fees from the leasing SaaS customs. In other words, we are aim to use a reliable and robust model and make the right pricing strategies after the evaluation of workloads cost accurately.This paper make some progresses facing the previous key problems and challenges. The main contributions of the paper as follows:we proposed a hybrid cloud storage op-timized framework——CloudMW, CloudMW can bring out several improved schemas to address the lack of performance fluctuation, data sharing support. By importing chunking and replication thinking of RAID into cloud storage, we stripe the data into clouds and u-tilize the middleware technology to implement data sharing and performance optimizing. The evaluation proved that CloudMW can avoid the performance fluctuation, support data sharing and maintain the online quality of service.In addition, the cloud storage reliability framework, called CloudRAID is presented. It is tendency that replace multiply replications to erasure coding as the efficient redundancy strategies in cloud storage platforms. Our work considered workload characteristics into the selection of cloud reliability schemas. After adopting the quantitative evaluation models, CloudRAID can group right reliability mechanisms adaptively. We practised CloudRAID in HDFS to validate the capability of service and effectiveness.Then, we put forward an adaptive I/O resource scheduling algorithm (Called CloudIO) for utility optimization to take the cloud providers’profit margin into consideration. Un-der the service capability guarantee of the premise of cloud storage provider, the algorithm can avoid the performance decreasing or stagnating and satisfy the Service Level Objec-tive(SLO). The results of our experiments suggest that the algorithm is adaptive to archive better workloads balancing while maximizing profits compared with other algorithms.We finally proposed the CloudCOST model for workloads cost in cloud computing environments. CloudCOST is of strong flexibility and not sensitive to the application con-solidation. We brought out workloads burstiness filter here:migrate bursty workloads to the clouds, where IaaS providers exist, and then let the rest processed locally. We ensure that our model is able to adjust itself according by the market changing. Through the validation by adopting of the actual IT company workloads, we found that CloudCOST is robust and make the right pricing mechanism synchronously.Through the above study about quality of service in cloud storage, this thesis is to be able to provide customers with excellent performance, high reliability, load balance and save the cost in cloud storage ecosystem at the same time.