Browse Prior Art Database

Efficient management of journal entries in a highly available messaging system

IP.com Disclosure Number: IPCOM000181868D
Original Publication Date: 2009-Apr-16
Included in the Prior Art Database: 2009-Apr-16
Document File: 2 page(s) / 57K

Publishing Venue

IBM

Abstract

Efficient management of journal entries in a highly available messaging system by dual key indexing

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

Page 1 of 2

Efficient management of journal entries in a highly available messaging system

Messaging systems commonly use journals to persist message and/or transactional data to disk, in the event of a system failure entries from these journal can be replayed to restore data/transactional integrity. For a highly available system, to remove a single system as a point of failure,

journal entries can be replicated to one

or more backup systems so that the replay may be performed on any one of the backups. The remote journal copies, such as those available in the IBM i5/OS * operating system are read-only at the backup system to guarantee integrity. One problem that is encountered is that replay on a backup instance may need to write persistent information to the journal, for example, to mark the resolved outcome of a transaction. One solution to this is to provide a seperate local journal for the new entries and to mirror this back to the production system - for example;

Asynchronous Replication

Synchronous Replication

    The messaging system needs to manage this information effectively at both production and backup systems. In the case of failover (production -> backup) and switchback (backup -> production) this may cause journal information to be located in segmented blocks in the local and remote journal receivers. The more often failover and switchback occur the more segmented the data becomes and locating an entry within the two journal chains becomes compoundly difficult and time consuming which is not ideal in a highly available environment.

When using journals,

journal entries are usually only retrievable by a system

Production System Journals

Local
(RW) Remote

(RO)

Backup System

Journals

Local
(RW) Remote

(RO)

Page 2 of 2

allocated journal sequence number and the application has no control over the allocation of this value, it is system generated - in addition this sequence number may be reset by the system which would consequently mean that a given sequence number is not unique. Journal entries can also be retrieved within a timestamp range, however there are drawbacks in that timestamp order may not be sequential in the case of a Daylight Savings Time transition and that for a high workload there may be hundreds of journal entries written with the same timestamp and so finding the right entry may require sequentially replaying a number of entries until the correct one is found.

    The system described herein proposes an efficient method to m...