Browse Prior Art Database

Procedure for Updating Replicated Directories

IP.com Disclosure Number: IPCOM000119940D
Original Publication Date: 1991-Mar-01
Included in the Prior Art Database: 2005-Apr-02
Document File: 3 page(s) / 114K

Publishing Venue

IBM

Related People

Austin, JH: AUTHOR [+4]

Abstract

Disclosed is a program design to allow replicated directory entries to be updated from multiple locations concurrently without mutual interference, permanent inconsistency, or multi-nodal locking.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Procedure for Updating Replicated Directories

      Disclosed is a program design to allow replicated
directory entries to be updated from multiple locations concurrently
without mutual interference, permanent inconsistency, or multi-nodal
locking.

      This procedure applies to a distributed database, such as a
directory, in which copies of individual records are maintained at
some or all nodes in the network, and queries are satisfied by
accessing any copy conveniently accessible to the requester.  This
procedure applies only to database records consisting of multiple
fields such that the value of any field does not depend on the value
of any other field or on the prior values of that field.  Although
this is not the most general class of distributed data, it is an
important subclass such as found in directories.  Examples of such
information are: a person's name, street address or telephone number.

      Benefits of this invention are that, given a distributed
directory whose individual entries may be replicated at some or all
nodes in a network:
   Update requests may be issued from any node or nodes in the
network, whether or not an instance of the entry exists at that node.
      Update requests issued from two or more different nodes at
about the same time will not interfere with each other.
      All replicas of the record will eventually contain identical
data.
      The effect of any requested change is immediately available to
the requester.
      It is not necessary to lock the data base at more than one node
at a time.

      For example, suppose a person at one node requests a change to
his telephone number, and, meanwhile at about the same time, an
administrator at another node requests a change to that person's job
title.  The directory service, using the technique described, will
cause all replicas of the record for that person throughout the
network to contain both changes, and neither requester must wait for
the other request to be serviced.

      For each record in the directory, one instance of the record
will be designated the Master instance of that record.  Other
instances of the record are called shadows. The masters for different
records may be at different nodes.

      The update request may specify the names of one or more fields
in the record and new values for those fields.  The new values may
not depend on the values of other fields or on the prior values of
that field.  The update request may NOT specify or contain the names
or values of any fields that are not to be changed by that request.

      An update request may be initiated at any node in the network.
It is then routed to the node containing the Master instance of the
entry.

      At the node containing the master record, the update requests
are processed, in the order received, by replacing the current values
of fields in the master record with the new values of only those
fields supp...