Implementing a relational database by means of specialzed hardware

Abstract

New hardware is described which allows the rapid execution of queries demanding the joining of physically stored relations. The main feature of the hardware is a special store which can rapidly remember or recall data. This data might be pointers from one file to another, in which case the memory helps with queries on joins of files. Alternatively, the memory can help remove redundant data during projection, giving a considerable speed advantage over conventional hardware.

Acceleration of a content-based image-retrieval application on the RDISK cluster

“…Although the concept of such processing capable storage devices dates back to the 70s [3], it has drawn an increasing attention at the end of the 90s [17,10,12,1] thanks to the emergence of Storage Area Network (SAN) and Network Attached Storage (NAS). The basic idea of a smart disk is to take advantage of the often underused computing power that is available on disk controllers, in order to implement some of the search operations directly on data as they flow out of the disk.…”

A performance evaluation of four parallel join algorithms in a shared-nothing multiprocessor environment

“…To enhance the performance of certain operations, an array of bit vector filters [BABB79,VALD84] When the hash table for the inner relation has been completed, the process sends its filter to its scheduler. After the scheduler has received all the filters, it sends them to the processes responsible for producing the outer relation of July 10, 1992 the join.…”

“…We also considered how effectively the different join algorithms could utilize processors without disks. Finally, bit vector filtering techniques [BABB79,VALD84] were evaluated for each of the parallel join algorithms.…”

Improving distributed join efficiency with extended bloom filter operations

“…f k }, which hash elements of U to an integer in the range of [1, m]. The m bits are initially set to 0 in an empty bloom filter 1 . An element e is inserted into the bloom filter by setting all positions f i (e) of the bit array to 1.…”

“…Bloom filter based joins The first hash-based join algorithm has been described in [1]. The simple hash-join works as follows: Suppose we want to compute S T , where S is the smaller relation.…”