A spatial hash join algorithm suited for small buffer size

Abstract

In this paper, a new algorithm for spatial join operations is introduced. The so-called NRQB (No Replication with Quadtrees and Buckets Spatial Merge Join) enhances the original PBSM by partitioning the space according to the spatial distribution of the objects. In addition, a hash file is created for each input data set and used to enhance both the storage of and the access to the minimum bounding rectangles (MBR) of the respective set elements. The paper also presents a performance evaluation of the proposed algorithm relying on the results obtained by the execution of a series of test cases concerning different spatial join scenarios. In each test case, the response time of NRQB is compared with that of some well-known algorithms. The test cases were conducted with both synthetic and real data sets. The results showed that the new algorithm is best suited for smaller buffer sizes, which are typical of mobile devices and database systems for desktop computers.

A Rule-Based Optimizer for Spatial Join Algorithms

“…Tree Transversal (STT) [Brinkhoff et all, 1993] and Histogram-based Hash Stripped Join (HHSJ) [Fornari and Iochpe, 2004].…”

“…The Histogram-based Hash Stripped Join (HHSJ) [Fornari and Iochpe, 2004] has three main characteristics: the object descriptors are stored in a hash file organization; a bi dimensional histogram of object distribution defines the spatial extension of each partition; and, when it loads objects to memory, it divides the space by strips.…”

Query optimizer for spatial join operations

“…The Histogram-based Hash Stripped Join (HHSJ) [9] has three main characteristics: the object descriptors are stored in a hash file organization; a bidimentional histogram of object distribution defines the spatial extension of each partition; and finally, when it loads objects to memory, it divides the space using strips.…”

“…In the number of I/O operations for the ISSJ, expressed by (9) or (10), half of the operations are reads and half are writes, because the sorting algorithm must load objects into memory and write the sorted set to disk. All reads and writes are sequential.…”

“…The query optimizer evaluates costs from four different algorithms: Partition Based Spatial Merge Join (PBSM) [19], Iterative Stripped Spatial Join (ISSJ), Synchronized Tree Transversal (STT) [4] and Histogram-based Hash Stripped Join (HHSJ) [9]. These four algorithms were chosen because they represent different classes of join algorithms and file structure.…”