Why Flexibility Is Key to a Successful Data Migration
Data migration is almost universally dreaded by IT professionals undertaking complex, application-level consolidation or renewal projects. A recent survey Celona conducted amongst telecoms IT professionals showed that 93% of them were fearful of application-level migration. This might seem a surprisingly high number, but unfortunately there are sound reasons for this level of unease. According to Bloor Research , over 80% of data migration projects are not delivered on time or to budget. Bloor's Phil Howard explains that research has revealed that Forbes 2,000 companies already spend at least $5 billion per year on migrations, and yet cost overruns average 30% and time overruns average 41%.
You might ask how this is possible? The answer is largely due to the fact that data migration is still not regarded as a valued skill and practice in its own right, but is treated as the final hurdle in a complex project - an afterthought once the functionality has been developed. At a recent British Computer Society (BCS) meeting BT's CIO Phil Dance bemoaned this fact. "Data migration is the single biggest thing that kills you," he warned. "You spend lots of money and effort on building functionality, but then you don't spend the effort on moving the data across…and now moving data isn't a once-in-a-career event, it's continuous. And it's even worse than that because now [we're a 24x7 company] we have no time to do it in."
So what makes one migration work where another fails to get out of the box? What ensures one comes in on budget, while another is massively overspent? Well, a successful migration is founded on getting three things right: people, process and technology. Getting any one or more of these three wrong will destabilise the entire project. That doesn't sound too difficult does it? Get the right experience, select a good method and then develop or buy a tool to deliver the data for you. But the problem is that large migrations are extremely complex on a variety of levels and the sheer scale of them soon leads to challenges.
Let's consider one of these three key success factors - migration technology. Analysing different migration projects reveals there are five main types of data migration, as shown in Figure 1. However, it is unlikely that any one of these will entirely fit all the requirements of a complex project. In other words, the chances are that any programme of complex transformation will require that a range of approaches can be delivered, as a single project may move through a number of approaches over time or even combine approaches in parallel.
For example, an enterprise might initially decide to go with a 'Don't Migrate' approach in order to get a new customer service up and running without any delays. Some information may then be synchronised with existing systems (eg revenues written back to the old accounts receivable system). However, following the launch and trial with new customers, the business may need existing customers who take up the new service to be migrated with their old service information on an event-driven basis. Then, after the new systems have stabilised, the business might decide that incremental or even bulk-load strategies should be used to migrate more customers to the new systems, while at the same time continuing to migrate individual customers when they order a new service.
|Don't migrate||Only new data is loaded to the new applications. However, new data may be synchronised back to old systems||Immediate start, no baggage from the old world|
|Event-based||Migrate one business object at a time when stimulated by a business event (e.g., new service is sold or element provisioned)||Business risk reduced and earlier realisation of value|
|Incremental||Migration is driven by the business and delivered in small, incremental steps (e.g., market, region or service)||Business aligned, low-risk and earlier realisation of value|
|Bulk load ||Large blocks of data are migrated together but the business transitions later||Simpler migration project|
|Big bang ||Large blocks of data, business process and people are transitioned together||In a simple migration this is the simplest strategy for IT|
Figure 1. The five main types of data migration. (Source: Celona Technologies 2008.)
As we have seen, a complex project may require multiple data migration approaches to be used at different times or in combination, in order to deliver the results needed by the business. However, data migration tools - whether proprietary or third party - have tended to be built to deliver a particular type of data migration. For example, although they have been used for other types of migration, extract-transfer-load tools (ETL) are designed to bulk load data warehouses. Since most proprietary and third-party tools are designed specifically to handle a migration in a certain way, they are often not able to accommodate business changes that impact the project, or handle the scenario where the business requires more than one data migration approach to be used. Delivering a complex application migration requires flexibility in the data migration tool used, as it is not always possible to predict how things are going to change during the course of the project, or exactly which approaches will be needed at the start of the project.
In response to the demand for flexible, multi-purpose tools, a new generation of data migration tools is emerging, as shown in Figure 2.
Progressive migration tools are one of the third-generation of third-party tools that are coming onto the market. They employ a flexible strategy for application-level data migration, enabling different migration approaches to be used during the course of the project, or in combination, as is needed. The flexibility inherent in progressive migration tools means they support both project and business change, and deliver a lower risk, faster and business-driven migration. Key benefits of progressive migration include:
Progressive migration delivers against one of the three pillars of a successful migration, but it is also essential that the other two pillars (people and process) are not neglected.