Research And Implementation Of A Segmentation Hybrid Temporal Index Structure In Database

In the era where data are being produced over time and shared in an unprecedented pace, mining the information in the big data has become increasingly crucial. Temporal information is the natural and basic description for the development and changes of real-world objects, and almost everything has explicit or implicit temporal features. And the data produced by the real applications are generated over time, they have temporal attributes naturally. From an academic point of view, temporal data management has been the subject of extensive research. While the traditional snapshot databases always record the information in a given specific time, it is difficult to reflect the dynamic changes of real world sufficiently and accurately. The need for temporal data management and retrieval shows increasingly urgent in most modern database systems today.Temporal indexes provide an important way to accelerate query performance in temporal database. However, it is a pity that the current temporal indexes are established on the temporal attribute, it can’t support the variety of queries very well. In addition, with the characteristic of constant update of the data in temporal database, it is hard to take account of both the efficiency of query execution and the index construction as well as maintenance. To deal with the above problems, this thesis pays more attention to how efficiently establish the index to support the consumers’ variety of queries.Firstly, this thesis describes the significance of temporal data management. It analyses research background of this issue and related technologies, highlighting the necessity of in-depth study to temporal index technologies.Then, the thesis introduces the data structure of the traditional index technologies in detail, including the data structure of the traditional B+-Tree index, Timeline index proposed by SAP HANA, the thesis also introduces the realizations of temporal operations on the index in detail, such as insert, delete, and update and so on. What’s more, the thesis analyzes the efficiency of index construction and maintenance, and the limitation of the variety of queries. Some parallel processing technologies for index construction are also introduced in this thesis.Subsequently, the thesis analyzes the structure design about the temporal index, and proposes a novel segmentation hybrid index: SHB+-Tree index for temporal data. First the temporal data in temporal table deposited is separated to fragments according to the time order. In each segment, the hybrid index is constructed which is a combination of temporal index and object index, and the temporal data is shared by them. To this index structure, this thesis proposes four different kinds of query type, they are Temporal Query, Object Query, Complex Query and Query over Segments respectively. The thesis also introduces the implementation method of the four query execution by the SHB+-Tree index. Because of the advantages of both the temporal index and the object index which the SHB+-Tree has, it can efficiently support the four temporal operations. Furthermore, by employing the segmented storage strategies and bottom-up index construction approaches for every part of the hybrid index, it greatly improves the performance of construction and maintenance. In addition, the object index is focus on the query about the object attribute, and the temporal index is focus on the query about the temporal attribute, if the query involves both object and temporal dimension, users can search the whole SHB+-Tree Index.Finally, this thesis uses benchmark data sets as test data to demonstrate these key technologies specifically. The experiment results conducted on the large-scale sets indicate that the presented strategy can support the variety of the queries, greatly improves the performance of temporal queries.