Research On Software Defined Storage On Cloud Hybrid Storage Technique

With the development of the Internet industry,the Internet today is in an era of information explosion.IaaS(Infrastructure as a Service)are often used to deploy big data analytics and other types of applications.Object storage allows cloud providers to easily scale up their storage,and it ensures high availability of storage by backing up data copies.In the aspect of storage me-dia,SSD in performance,energy consumption and other aspects compared to HDD has greater advantage,because of its high price and limited capacity,it is still a long way to go before com-pletely using SSDs for enterprise data storage.Software defined storage is more suitable for hybrid storage system,because its autonomically optimizes architecture of storage cluster.WHOBBS[1]is our previous research on hybrid storage.WHOBBS,while effective,only supports single monitor architecture as it concentrates all the monitoring tasks on one single node.Thus,single monitor would be a serious problem for WHOBBS.In this thesis,we design and implement DOBBS,which extends WHOBBS with multi-monitor architecture and splits the cluster into subclusters.We put forward the concept of Local Heat Balancing for manag-ing the system efficiently.Local heat balancing,actually,indicates fully exploiting the unique performance potential of SSD,and ensures high accessed data reside on SSDs.We propose a concept named Global Heat Balancing which ensures the whole cluster is in the state of heat balance.Global Heat Balancing includes two parts,heat imbalance detection and heat diffusion.Based on system design,we implement the system and introduce the implementation details and the main workflows of each module of DOBBS.In order to verify the efficiency of DOBBS,we conducted a large number of experiments.A large number of experiments show the efficiency of DOBBS and the effectiveness of Global Heat Balancing.