Browse Prior Art Database

Reread During Internal Sort

IP.com Disclosure Number: IPCOM000078406D
Original Publication Date: 1972-Dec-01
Included in the Prior Art Database: 2005-Feb-25
Document File: 1 page(s) / 11K

Publishing Venue

IBM

Related People

Parkinson, J: AUTHOR

Abstract

During an internal sort a cycle is performed, in the first part of which one or more sequences are produced by sorting records in a record storage area. These sequences are then written on a work device. In the second part of the cycle, part of the record storage area is used as input buffers into which these sequences can be reread. These sequences are then reread and at the same time merged with a sequence being produced in the remaining part of the record storage area. The sequence resulting from this merge operation is an output sequence. This method uses I/O most efficiently writing and rereading one sequence per cycle.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 88% of the total text.

Page 1 of 1

Reread During Internal Sort

During an internal sort a cycle is performed, in the first part of which one or more sequences are produced by sorting records in a record storage area. These sequences are then written on a work device. In the second part of the cycle, part of the record storage area is used as input buffers into which these sequences can be reread. These sequences are then reread and at the same time merged with a sequence being produced in the remaining part of the record storage area. The sequence resulting from this merge operation is an output sequence. This method uses I/O most efficiently writing and rereading one sequence per cycle.

A slight modification of the idea could be applied to polyphase. Given an n- tape system, one normally writes sequences with the same ordering on at most n-2 tapes, and then with reverse ordering on the same tapes, etc. Since one tape is not used during this process, it would be possible to write n-2 intermediate sequences, with the same ordering as the last set of output sequences. These would then be combined one at a time with n-2 second-phase sequences to the normal output tapes. Performance will depend on blocking factors, and almost certainly will be best with a smaller clocking factor for the intermediate sequences than for internal sort output; at least if the intermediate merge is to use all available ore without a great many extra buffers. Since in its use of data passed, the method is nearly as efficient...