Browse Prior Art Database

Method for Maintaining Persistent State on a Storage Medium

IP.com Disclosure Number: IPCOM000123770D
Original Publication Date: 1999-Apr-01
Included in the Prior Art Database: 2005-Apr-05
Document File: 2 page(s) / 70K

Publishing Venue

IBM

Related People

Bolam, SW: AUTHOR [+5]

Abstract

Any system which has to store a large amount of persistent state will try to minimize the amount of data needing to be processed on restart, e.g. by checkpointing using a serial medium such as a log. When checkpointing is applied to a medium providing both serial and direct access the following method can be used to minimize the amount of "change" records which need to be processed on restart.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 53% of the total text.

Method for Maintaining Persistent State on a Storage Medium

   Any system which has to store a large amount of persistent
state will try to minimize the amount of data needing to be
processed on restart, e.g. by checkpointing using a serial medium
such as a log.  When checkpointing is applied to a medium providing
both serial and direct access the following method can be used to
minimize the amount of "change" records which need to be processed on
restart.

   Consider a system which needs to store state for a large
number of entities.  Entities can be created, changed and deleted.
Using a direct access medium each state change could be made
persistent by: (1) creating a new entity by writing a new record, (2)
changing the state stored for an existing entity by deleting its
record and replacing it with a new record, or (3) deleting an
existing entity by deleting its record.

   In the above method N records have to be read on restart
(N=number of entities).  When the restore time is proportional to N
this method is not viable.  To reduce the number of records which
need to be written a threshold T can be introduced at which time the
state is checkpointed.  This involves deleting all records and
replacing them with one or more records each containing a large
number of entities.  Using this method, further state changes can be
made persistent by: (1) creating a new entity by writing a new
record, (2) changing the state stored for an existing entity by
writing a new record containing the new state, or (3) deleting an
existing entity by writing a new record indicating that the entity
had been deleted.

   This last method uses the direct access medium like a
traditional log - on restart the records are read serially to
restore the state.  To use the direct access capabilities of the
m...