Browse Prior Art Database

Asynchronous Recovery And Reliability for a Multiuser Database System

IP.com Disclosure Number: IPCOM000107776D
Original Publication Date: 1992-Mar-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 3 page(s) / 125K

Publishing Venue

IBM

Related People

Sadjadi, S: AUTHOR [+2]

Abstract

Disclosed is a design for achieving software recovery and reliability asynchronously in a multiuser/multiprocess database system. Primary objectives include providing data integrity, failure recovery, early error detection, structural repair, structural integrity verification, continuous processing and global resource management -- all accomplished with minimal performance impact.

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

Asynchronous Recovery And Reliability for a Multiuser Database System

       Disclosed is a design for achieving software recovery and
reliability asynchronously in a multiuser/multiprocess database
system.  Primary objectives include providing data integrity, failure
recovery, early error detection, structural repair, structural
integrity verification, continuous processing and global resource
management -- all accomplished with minimal performance impact.

      Recovery and reliability are especially important concepts
concerning a database management system (DBMS).  In a multiuser
database environment it is essential to provide continuous processing
and the ability to recover and repair damaged data structures.  This
entails including a "recovery manager" to provide functional recovery
and cleanup on behalf of failed processes.  Currently, functional
recovery is not inherently supported by the operating system or
database management system.  Following process failure, it is the job
of the system administrator or programmer to perform tasks such as
structural integrity verification, process failure detection and
recovery, damage control and recovery, failure causal analysis, and
point-of-failure recording.  A more desired approach, however, is to
have a recovery manager provide the same functions asynchronously,
without halting the system due to the failure of one component.

      The figure illustrates the five components of the proposed
database design.  The Resource Manager is the parent process for the
system.  It starts the other processes and allocates all shared
resources which are accessed through the Control Blocks.  Database
management functions are performed by the Database Engine which sends
any page- level I/O requests to the I/O Subsystem.  The I/O Subsystem
is made up of an I/O Scheduler which routes requests to concurrently
running I/O Servers.  A Recovery Daemon or Recovery Manager monitors
the system and provides asynchronous recovery and reliability.

      The Resource Manager allocates shared memory (resources), based
on configuration information, on behalf of all processes using shared
resources.  It then initializes related pointers in the Global
Control Block to corresponding resource addresses, maintaining
relative offset positions from the beginning of each shared memory
segment (e.g., the Buffer Manager pointer will be initiali...