Availability through a migration

Zero-downtime migration can prevent system failures, Steve Roy points out

The 2008 opening of London Heathrow's Terminal Five (T5) is a day many people will find hard to forget: the baggage chaos that ensued due to the faulty baggage handling system meant many irate travellers suffered cancelled and delayed flights.

Although IT staff had tested the system before T5’s opening, glitches in the baggage handling system are said to have cost British Airways, the only airline to fly from the new terminal, approximately £16m.

Business continuity really is vital – particularly in the case of business-critical environments. Migration processes – for databases, applications and hardware environments, for example – are complex tasks that can take months if not years.

We assert that the industry has been plagued with projects that have gone awry due to bad strategy and planning.

Even if the data component is not the biggest aspect of a migration project, it is still the most complex – and usually the step that determines project success.

According to a study carried out by IT market analyst Bloor, 13 per cent of all migration projects are completed on time and budget.

There are many reasons for this – from lack of experience, incorrect configuration and overconfidence right through to plain wishful thinking!

The oft-used ‘big bang’ way of commissioning new systems is risky. System changeovers are often done over a weekend in the hope that, come Monday, the new system will start without a problem.

Yet companies doing this are also risking losing everything all at once.

Many companies panic at the thought of a migration project, particularly where large volumes of data are concerned. Postpone the migration, though, and the company may need to migrate not from version x to x+1, but to a more complex version x+n.

If there is a valid business reason for migration to version x+1, technical considerations should not stand in the way.

So, we think, companies need a clear and reproducible methodology that will make such projects seem less daunting.

Application and database upgrades -- such as upgrading from Oracle 8i to 10g or Siebel 6.2 to Siebel 8 -- should ideally be interruption-free.

Some organisations may choose to deploy a zero-downtime solution.

The old system and new system run in parallel, enabled by real-time bi-directional data movement and synchronisation between the primary and secondary underlying databases.

Both can process transactions simultaneously and at the same time handle any data collisions or conflicts, resolving them via pre-defined business rules.

End-users get continuous availability. If one system goes down, the other can take the full processing load without even having to alert the end-user.

Having fail-back or contingency during a migration phase is critical to continuous uptime.

Irrespective of the migration method or product chosen, start planning in good time.
Many issues need to be clarified, such as: where is the data coming from that needs to be migrated? What is the source database? Which data is to be migrated? What are the characteristics of this data?

Asking these questions can clarify whether ‘big bang’ or zero-downtime migration is most suitable. Many providers and solutions on the market offer the ‘big bang’ method but fewer offer continuous availability.

Steve Roy is sales director at GoldenGate Software