On the SD-tree construction for optimal signature operations

Abstract

Signature file based access methods initially applied on text have now been used to handle set-oriented queries in ObjectOriented Data Bases (OODB). All the proposed methods use either efficient search method or tree based intermediate data structure to filter data objects matching the query. Use of search techniques retrieves the objects by sequentially comparing the positions of 1s in it. Such methods take longer retrieval time. On the other hand tree based structures traverse multiple paths making comparison process tedious. In this paper we describe a new indexing technique for OODB using the dynamic balancing of B+ tree called SD (Signature Declustering) tree. In this work the positions of 1s in the signatures are distributed over a set of leaf nodes. Using this for a given query signature all the matching signatures can be retrieved cumulatively in a single node. Also for signature insertion and query searching an optimal search path is calculated so that the entire process is speeded up. Experiments have been conducted to analyze the time and space overhead of the SD-tree by varying the signature length and the distribution of signature weight. The study clearly indicates the advantage of fast retrieval time in a way quite different from the other methods suggested in the past.

Một Số Cải Tiến Cho Hệ Truy Vấn Ảnh Dựa Trên Cây S-Tree

“…Elizabeth Shanthi và cộng sự [24] đã tiếp cận phương pháp truy vấn dữ liệu dựa trên chữ ký nhị phân bằng các cấu trúc dữ liệu khác nhau như SSF (Sequential Signature File), BSSF (Bit-Sliced Signature File), CBSSF (Compressed Bit-Sliced Signature File), S-Tree, SD-Tree,… Trong bài báo này đã thực hiện các thao tác chèn, tìm kiếm và xóa chữ ký trên cây SD-Tree. Trong thực nghiệm của bài báo này đã cho thấy phương pháp truy vấn trên cây SD-Tree cải thiện đáng kể tốc độ truy vấn.…”

A Signature-based Grid Index Design for RFID Main-Memory Databases

“…A selected set of the attributes of a piece of data is processed and stored as bit vectors, referred to as signatures, for efficient identifying of the existence of a piece of data in the database. The closest research is on the signature-based techniques proposed for object-oriented database systems [20,4,16,24]. Other related work includes the response-time enhance methods based on index structures, such as S-Tree [7], and their derivatives [25,26].…”