Browse Prior Art Database

Directory Organization for a Storage System

IP.com Disclosure Number: IPCOM000094661D
Original Publication Date: 1965-Apr-01
Included in the Prior Art Database: 2005-Mar-06
Document File: 2 page(s) / 32K

Publishing Venue

IBM

Related People

Brenza, JG: AUTHOR [+3]

Abstract

The drawing shows a directory organization for a content addressable memory system. In such a system, one index is assigned to the directory for each independently addressable record. For a very large storage system, the directory then grows to a considerable size itself. In addition, the directory contains very high traffic information, since all or part of the entries in the directory must be searched for each file data reference. It is therefore desirable to reduce both the quantity of stored information and the search time in the directory.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Directory Organization for a Storage System

The drawing shows a directory organization for a content addressable memory system. In such a system, one index is assigned to the directory for each independently addressable record. For a very large storage system, the directory then grows to a considerable size itself. In addition, the directory contains very high traffic information, since all or part of the entries in the directory must be searched for each file data reference. It is therefore desirable to reduce both the quantity of stored information and the search time in the directory.

This is accomplished in this directory by applying the record name which may, for example, be N-bits long to a randomizer 10. The output from randomizer 10 is a bucket address and a tag. The sum total of the bits required to make up the bucket address and the tag is considerably less than that required for the original record. The number of bits B+T is set large enough so that the duplication mapping problem is set to a tolerable maximum. A first memory 12 is provided which has a number of bucket addresses, each of which contains a plurality of storage positions. The bucket address portion of the output from randomizer 10 is applied to address one of the bucket positions in memory 12. The tag portion of the output from randomizer 10 is then stored in memory 12 at the first position in this bucket address. All other entries at this bucket address are moved down one position. The fact that the bucket address does not have to be stored further reduces the amo...