Research On Long-term Storage Cost Optimization Strategy Of Scientific Workflow Based On Cloud Computing

At present,the cloud computing platform mode is developing and improving rapidly,and the computing and storage capacity of the system has been greatly improved.With the advent of the era of big data,no matter the scale of data in the financial industry or the increase of communication between academic circles,the construction of cloud computing platform is put on the agenda.For example,in the research of Electric Power Patrol,many countries adopt the way of combining UAV to establish a perfect scientific workflow platform based on cloud computing,which can quickly,conveniently and accurately achieve the established goals.However,in the process of scientific workflow running for a long time,a large number of useful intermediate data will inevitably be generated.How to deal with and reduce the cost of the cycle has become a hot topic for scientists.Based on this background,our thesis has made work in the following two aspects:First,a long-term data storage optimization strategy based on multi-objective is proposed.Because the data in scientific workflow is reused or shared with each other,the system needs to select some key data storage for future use.But in the long-term storage process,because of the large amount of intermediate data,only relying on the massive storage capacity of cloud computing platform will lead to high storage costs.Therefore,after determining the storage nodes,our thesis will use the strategy of tiered storage to reduce the long-term storage costs.The specific strategy is to tier the node data to be stored according to the frequency of use in the cycle,place the high frequency in the high-speed storage,common data to conventional treatment,low frequency data archiving.At the same time,pay attention to the extra delay time.Experimental results show that this strategy can solve the problem of high cost caused by a large number of data to be stored in long-term storage,and can guarantee the user's delay demand.Secondly,a time delay strategy for dynamic determination and allocation of key nodes is proposed.The establishment of this model not only needs to meet the independent needs of users,but also needs to analyze the importance of workflow nodes.The potential cost of multiple searches caused by changes in price parameters in long-term stored procedures,so the model in our thesis also needs to achieve the goal of dynamic measurement,and can trade-off between fast and accuracy.The experimental results can verify that compared with the accurate algorithm,the cost of our model is slightly improved,but the time complexity is greatly reduced,it can adapt to the dynamic scientific workflow background,and improve the user satisfaction.To sum up,our thesis is based on the scientific workflow scenario of long-term storage,and focuses on solving the high cost problem in long-term environment.Through the idea of hierarchical storage,to ensure the user's deadline demand and reduce the cost.By analyzing the specific scientific workflow,we can find the key nodes dynamically and quickly,allocate the deadline according to the weight,and minimize the cost of the whole long-term storage workflow based on the cloud computing environment.