You'll never love an appliance like your old database

What stops you committing to the all-in-one hardware solution?


A more specialised appliance

Perhaps this explains the gradual move from database appliances as generic boxes designed to run everything towards a more workload-specific product.

"When Oracle first introduced its database appliance, it envisioned the box as a general-purpose unit that could run multiple database applications, serving a variety of needs. The company put forward consolidation of various small databases as a use case. In practice, people don’t use database appliances in that way," Vile suggests. “They found their niche around single-purpose, single-workload, high-throughput applications."

These will often revolve around specific workloads. “The appliances will become analytical appliances rather than database appliances,” according to Quocirca’s Clive Longbottom.

This category of devices has been evolving for a while, and you’ll find data warehousing appliances from the likes of IBM, Teradata and HP. As companies’ data landscapes become more varied, though, Longbottom sees the use case transforming yet again.

“This was a case of throw us your high-value data, which you throw us in near-real time, and we’ll do it on steroids,” Longbottom adds. That was fine when 80 per cent of a firm’s data was relational and 20 per cent was non-relational, he argues, but that statistic is inverting.

This movement raises other issues for companies exploring the database appliance space. More data is moving into non-relational forms, meaning companies need a combination of database functionality that can handle data in a variety of formats from multiple sources.

This has created what Longbottom calls a “three-legged stool”, containing something that can serve as a data filter, which is Hadoop, and the relational side, and then the non-relational processing layer.

Some vendors are putting elements of all these things into the same box. Pivotal has an Hadoop-based environment that is expandable in modular form, with the ability to add processing capabilities for structured and unstructured data. Teradata has its Aster big analytics appliance.

What does all of this mean for the traditional discrete database, running on a box purchased from a separate vendor? It’s unlikely to go away. Companies are likely to still resort to dedicated databases for many operational applications, especially when their workload is difficult to size, or where an IT team is confident enough to configure its own build to retain more flexibility.

“Another layer is master data management and the database management within your wider organization,” says Rackspace’s Fox. “That isn’t necessarily solved by appliances, but by the vast plethora of options in the vast data centre management space.”

Database appliances were introduced to simplify data management for a layer of mid-management companies with uncomplicated needs. But they were introduced just at a time where things were getting more complex. Several years ago, companies relied on a handful of large vendors for their processing needs, but things have changed.

With a greater array of data processing requirements and more vendors all offering solutions for specific forms of data management, it’s unlikely that generic database appliances will commoditise discrete databases out of existence. ®


Other stories you might like

Biting the hand that feeds IT © 1998–2022