Research Of Dynamic Skyline Query Processing Approach In MapReduce

Recently,with the rapid development of information technology and widespread application of Internet,the global data storage is showing the scale of explosive growth.So how to find interesting data objects in large-scale data set has become the focus in the study of database.Skyline query can compute the optimal solution which meets multiple standards in large-scale data.It has been widely applied for multi-objective decisions.The traditional skyline query is static,it only consider the static attribute value of each dimension of data points.As long as given a data set,a skyline query result is certain.Dynamic skyline query,as an important variant of skyline,its result can be dynamically changed with choosing different query point,providing more flexibility when the users make some specified needs.However,dynamic skyline query can return a large number of redundant data and ignore the directionality of query point and data integrality,lacking of more global optimal results,making it difficult for users to choose.It is necessary to optimize the result set of dynamic skyline,improving the whole data integrality and filtering a large number of redundant data.In order to effectively improve query efficiency in the massive data,the thesis studies the dynamic skyline query processing method in MapReduce.Firstly,on the issue of the result set of dynamic skyline has a large number of redundant data and lacks more global optimal results,we propose augmented dynamic skyline in MapReduce.It can combine with the characteristics of global data distribution.The algorithm partitions the original data according to dimensional information using q point(query point),parallel computes dynamic skyline points in multiple nodes,optimizes the result set of the traditional dynamic skyline and at the same time provides more global optimal results for the user to choose,the data integrality has been increased.It reduces the comparisons of meaningless dominance relationship between data points.In this case it reduces the computing time and storage space overhead.Secondly,users may consider user preferences.For some attribute,they have own requirements in practical application.In order to fully consider the case of user preferences,we propose user tolerances constrained augmented dynamic skyline in MapReduce.This method builds dimension attribute index in each dimension.Then it reduces the original data set according to the user tolerances.Finally it computes the augmented dynamic skyline.This method reduces the original data set according to the user tolerances and the dimension attribute indexes.In this case it reduces the size of result set and the comparisons of dominance relationship between data points.It reduces the computational overhead and the response time.The results meet the needs of users still further.In the case of user tolerance can lead the reduced original data sets to be empty,we propose tolerance pointer down strategy based on dimension priority.It eases tolerance in the dimension of the low priority,and gives users alternative points,then computes augmented dynamic skyline.Finally,in order to verify the validity and rationality of this algorithm,we have conducted the experimental analysis from four aspects of the query response time,the size of result set,the comparisons of dominance relationship between data points and the influence of user tolerance to the size of the result set of MR-T-ADS.A large experiments show that our proposed dynamic skyline query processing method in MapReduce can reduce a large number of redundant data and improve the query efficiency.