Research On Storage And Optimization Of Uncertain Time Series In Relational Databases

As a common form of data existence,Time-Series is widely used in many fields in the real world.Through processing the Time-Series data which were observed and collected,people may solve many problems such as decision-making and prediction.Thus it is of great significance for the macro-control of the national economy,weather forecasting and earthquake precursor prediction.Because of the ubiquity of uncertain data,the theory of storage and management based on deterministic Time-Series data are not suitable for reality needs now.Firstly,traditional Time-Series databases generally use columnar storage in a distributed environment.Storage of uncertain Time-Series data in this form may lead to waste of storage resources.Secondly,compared with crisp data,uncertain data corresponds to multiple data at one time point.So using traditional sequential database storage is more possible to cause data confusion.At present,there are few research works about the storage of uncertain time series data were published.This thesis is devoted to the research on the storage method of uncertain TimeSeries on relational database.On the basis of overcoming the defects of traditional database,this thesis attempts to find a convenient data storage method to realize the effective storage of uncertain TimeSeries data.The main work of the thesis includes two points:Firstly,the An uncertain Time-Series storage method based on relational database is proposed.Firstly,the unstructured uncertain Time-Series is described by the traditional semi-structured language,and the problems caused by the storage to the relational database are analyzed.Secondly,by analyzing these problems and combining the storage characteristics of traditional time series database,a framework for uncertain Time-Series storage is proposed.The storage model is then designed at the relational database logical layer and the best data organization is selected at the physical layer to efficiently store the indeterminate time series.Finally,a prototype system for storing uncertain TimeSeries is designed and developed,and the storage structure is analyzed and compared to verify the feasibility of the proposed storage method.Secondly,a method for optimizing an uncertain Time-Series storage system is proposed.Write efficiency and query efficiency are important indicators for uncertain Time-Series storage.In order to improve the write rate and query efficiency of the storage,a double cache mechanism is established for the uncertain Time-Series characteristics to separate the read and write operations and achieve certain operations.In the case of volume,the I/O resources requested by the system are reduced by triggering the submission strategy,thereby improving the writing efficiency.In addition,the hot data selection strategy is used to put the hot uncertain time series data into the query cache to improve the query efficiency.The experimental results show that compared with the storage system without the cache mechanism,the proposed storage prototype system optimization method with added cache mechanism has higher efficiency in writing and querying.