Browse Prior Art Database

Sorting With Cached DASD

IP.com Disclosure Number: IPCOM000047114D
Original Publication Date: 1983-Sep-01
Included in the Prior Art Database: 2005-Feb-07
Document File: 1 page(s) / 11K

Publishing Venue

IBM

Related People

Bussey, DR: AUTHOR [+4]

Abstract

This article describes a program implemented method for sorting data utilizing the buffer of a cached DASD (direct-access storage device) as staging device. The cache controller is supplied with a list of tracks or blocks in the order in which they will be needed; the tracks or blocks are prefetched from a sort work file on DASD and bound into the controller buffer (cache) until the buffer portion allocated to sort is filled, and each track or block is marked for release from the buffer when no longer needed. The prefetching of tracks or blocks continues until the sort work file is empty. This method saves disk latency (seek and rotational delay) time. Asynchronous write allows the channel to return device-end as soon as data being written to the sort work file is in the buffer, and not wait until the data is committed to DASD.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 95% of the total text.

Page 1 of 1

Sorting With Cached DASD

This article describes a program implemented method for sorting data utilizing the buffer of a cached DASD (direct-access storage device) as staging device. The cache controller is supplied with a list of tracks or blocks in the order in which they will be needed; the tracks or blocks are prefetched from a sort work file on DASD and bound into the controller buffer (cache) until the buffer portion allocated to sort is filled, and each track or block is marked for release from the buffer when no longer needed. The prefetching of tracks or blocks continues until the sort work file is empty. This method saves disk latency (seek and rotational delay) time. Asynchronous write allows the channel to return device-end as soon as data being written to the sort work file is in the buffer, and not wait until the data is committed to DASD. Since the sort work file is only a temporary storage file, with asynchronous write there may be some situations where it is not necessary to ever write to DASD. Further, the use of block level (rather than track level) bind and release will result in elapsed sort time savings since the data being staged would be more closely related to the sort program and, consequently, more likely to be staged in the appropriate order. As an alternative to pre-fetching tracks into the buffer, the track may be staged at its first reference. This would result in some delay to transmit the entire track into the buffer on its first r...