ERROS - High Availability Organiser

The IBM iSeries (AS/400,i5, Power Systems i) is a very robust server platform on which to run business applications. It has single level storage, an inbuilt relational database, inbuilt communications and security software, as well as built in Apache and HTTP servers. These are all part of and totally integrated with the extremely stable operating system (OS/400, i5/OS). It can have mirrored disk sets or hot swappable disk drives in Raid sets. It can have multiple Ethernet and other communications links to allow multiple communications route to the server.

ERROS, with its unique design and resulting simplicity, is also very reliable. Together, the IBM iSeries (AS/400, i5, Power Systems i) and ERROS make a most stable operating environment, but things still can go wrong - there can be lengthy power cuts, communications failures or natural or other disasters.

An option of ERROS is the ERROS High Availability Organiser (HA Organiser). This allows replication of all ERROS transactions on the prime system on up to 255 local and/or remote servers. These can be at differing levels of OS/400. Each target system can, in turn, be replicated on 255 further local or remote servers, so that there is no effective limit to the number of copies of the database that can be replicated and kept up to date. The remote target systems can be anywhere in the world. Subject to any communications problems, the ERROS HA Organiser ensures that the remote copies of the ERROS database are updated continuously and almost instantaneously. The ERROS HA Organiser works with all existing and future ERROS applications, whether these are created by ERROS Ltd or by an ERROS licensee. It will not provide high availabillity for non-ERROS applications.

Not only does the ERROS HA Organiser provide almost immediate backup for the prime system, it allows part of the processing workload to be transferred to a target system so that the prime system's processing and communications capacities can be reserved for transaction processing and database updating only. Any users of any ERROS applications or ERROS based websites who do not require to update the database can use a target system for their enquiries and read only tasks. Whilst use of the ERROS HA Organiser on a target system is an effective backup of all ERROS application data and programs, full off-line backups of the prime system must still be taken regularly. Whilst this is being done, the target system can still be used for enquiries and read only work. Once the prime system has had a full backup and has been restarted, then the target system should also be backed up off-line. When the target system is has been restarted, the ERROS HA Organiser will ensure that the target system is brought up to date as quickly as possible. Because all ERROS files are always journalled, a full off-line backup only needs to be taken weekly. A standard ERROS facility allows automatic hourly off-line backups of all changes to the ERROS database and to ERROS programs since the last full backup and backup of all changed PC stream files on the Integrated Filing System (IFS) although the latter are not currently replicated by the ERROS HA Organiser. This automatic hourly backup takes place whilst the system is in use.

Since ERROS applications are largely defined in the ERROS database rather than in programs and the ERROS kernel programs change rarely, most, and often all, changes to an existing ERROS application, including those to the structure of the data, as well as any newly created ERROS applications, will be replicated automatically to the target systems. As ERROS authorities to those applications are also stored in the ERROS database, these also will be replicated to the target system(s). If, due to commitment control, a rollback occurs on the prime system, the resulting journal entries will be processed to rollback the same entries on target systems.

In most computing environments, planning and implementing a high availability solution is a complex task. Setting up the prime system and any target systems to take advantage of the ERROS HA Organiser is very straightforward. As all files in the ERROS database are always journalled and the ERROS HA Organiser uses the Remote Journalling facility of the iSeries which allows all database changes to files that are journalled to be copied to target servers, no new programs or program changes are required for the prime system. If you have a second iSeries available for use as a target system, and your prime system can communicate with that machine, then setting up the ERROS HA Organiser should take well under 1/2 hour. You will of course have to save your ERROS database and ERROS programs on your prime system and restore them to the target system before you can implement the ERROS HA Organiser.

Because of the inherent simplicity of the structure of all ERROS applications, the ERROS HA Organiser is also extraordinarily simple and consists of just two programs which run on ERROS target systems - one to receive the journal entries and the other to apply the changes to the ERROS database on the target system. Because Remote Journalling is part of the iSeries Operating System, it uses very little processing power on the prime system. The structure of ERROS ensures that only those attribute iterations that have changed are journalled, rather than a complete record. This means that journal receiver entries are much smaller than might otherwise be the case, considerably reducing the amount of data to be transmitted to the target machine. As a result, applying the journal receiver entries on the target machine does not consume high system resources and is most efficient. Administration of the ERROS HA Organiser is very simple, particularly when compared with other HA products.

The ERROS HA organiser does not have an automatic switchover facility if the prime system fails. The iSeries can be set up to handle some points of failure automatically - for instance, if a network adapter should fail, using the Virtual IP address of OS/400 (V5R2 and later), it could automatically switch to another adapter. However, once you have decided which target system (if you have more than one) is to become the prime system, making the switchover manually is a simple, speedy task, taking only a few minutes. No data restoration is required as the target system will already be up to date, unless the prime system failed to complete and transmit a transaction. All ERROS applications on target systems are automatically read only. When the target system is designated as the prime system, all ERROS applications immediately allow update, without any changes to those applications or their authority being necessary. Communication and transmission of the remote journal receivers can be done synchronously or asynchronously. If transmission of the remote journal receiver to one, or the only, target system, is synchronous, then that target system will always be totally up to date as it will be updated before the prime system.