Research And Implementation Of Key Technologies Of Cloud Storage Scheduling

With the popularity of the Internet in various industries, and the use of big data, the amount of data that needs to be stored is in a rapid increase. This trend prompted the rapid development of cloud storage which is cheap, safe and has high capacity. Cloud storage is a model which stores data in the logical storage pools. Users need not to care about the specific physical implementation. In this mode, the physical storage device is generally provided by a cloud storage service provider. Service providers are responsible for availability and security. Users only need to pay for storage services, do not need to care about other details.Cloud computing is first proposed by Amazon as a public cloud platform, then with the development of the open source cloud platform, cloud storage users begin to build their own private cloud. The situation that public cloud and private cloud have their own characteristics prompts the emergence of hybrid clouds. Hybrid cloud combine public and private cloud advantages, has been widely recognized, and is considered the future direction of cloud storage. However, since the structure is more complex hybrid cloud compared to the other two cloud storage mode, so there are some hybrid cloud problems to be solved.Cinder is the block storage module for Open Stack, which is responsible for providing storage volumes for virtual machines. Open Stack has been widely used and brought Cinder used more and more widely. However there are some flaws in Cinder, it is necessary to improve them.In this thesis, we will propose new methods to solve hybrid cloud storage scheduling and Cinder scheduling. The main contents are as follows:(1) Hybrid cloud storage has the advantages of public cloud storage : low-cost, easy to expand and the advantages of private cloud storage- safe and efficient. But the public and private clouds binding also increases the difficulty of scheduling about hybrid cloud. Most current research is only scheduled for a single target such as load balancing, access speed, price and so on. However, no scheduling try to take multiple aspects together into account, such as security which is special in hybrid cloud. In this paper, we will consider various factors then propose a scheduling algorithm, which is based on BP neural network.(2) Cinder only considers scheduling the node which has the largest remaining storage capacity. This method does not has the ability to get node which has the best overall performance. Therefore, we propose a scheduling method based on triangular fuzzy number. This method can considers multiple evaluation factors. At the same time the introduction of triangular fuzzy numbers leads to qualitative indicators can be described using quantitative methods.(3) Finally, we designed a cloud platform which is based on Open Stack. We use our scheduling algorithms in the storage module of the platform. We find that the hybrid cloud scheduling and the multi-node scheduling overall performance are improved.