Browse Prior Art Database

Dual Striping for Replicated Data Disk Array

IP.com Disclosure Number: IPCOM000103798D
Original Publication Date: 1993-Jan-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 1 page(s) / 66K

Publishing Venue

IBM

Related People

Merchant, A: AUTHOR [+2]

Abstract

In this disclosure, a dual striping scheme, which is a data declustering and striping scheme for disk arrays with replicated data, is proposed to provide improved performance in query and transaction processing systems. Generally speaking, arrays of disks with replicated data are used to provide reliable, large data storage with high throughput at low cost. There are several ways of partitioning, or striping the data and the copy. In this disclosure, it is suggested that, by maintaining the two copies in stripes of different sizes, and keeping both copies actively accessible at the same time, substantial improvements in throughput and delay characteristics may be obtained in the context of database applications.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 61% of the total text.

Dual Striping for Replicated Data Disk Array

      In this disclosure, a dual striping scheme, which is a data
declustering and striping scheme for disk arrays with replicated
data, is proposed to provide improved performance in query and
transaction processing systems.  Generally speaking, arrays of disks
with replicated data are used to provide reliable, large data storage
with high throughput at low cost.  There are several ways of
partitioning, or striping the data and the copy.  In this disclosure,
it is suggested that, by maintaining the two copies in stripes of
different sizes, and keeping both copies actively accessible at the
same time, substantial improvements in throughput and delay
characteristics may be obtained in the context of database
applications.

      In this scheme, one copy of the data is range partitioned into
large stripes, one per disk in the disk array.  The copy of each such
stripe is broken into page-sized pieces, which are distributed over
the remaining disks.  All the disks are under a single controller,
and both copies of the data are actively accessible.  Requests to
read small quantities of data, say, one page, typically generated by
transactions, are directed by the controller to the small stripe
copy, whereas requests to read large amounts of data, typically due
to queries, are directed to the corresponding large stripe.

      If one disk malfunctions, the small page reads that would
normally have been read from this di...