Database Virtualization: Cloudy Expectations
Tuesday, June 02, 200909:47 am EDT
In James Kobielus’ recent blog (http://jkobielus.blogspot.com/2009/05/forrester-blog-repost-database.html) he points out that database virtualization has been around for a long time. Many of these database virtualization schemes are built on abstractions to simplify data access, but the emergence of system virtualization (aka the hypervisor) has raised the expectations for database virtualization. It’s no longer about whether an application can get to its diverse data, but about how fast and reliably that data can be delivered. On top of this increased demand of the database is the hype about cloud computing, which creates an even greater expectation that the database infrastructure should also be elastic(i.e. to be able to grow or shrink processing capacity on-the-fly through dynamic and automatic provisioning).
The advent of cloud computing has changed the focus from database virtualization to exactly what the database itself should look like in the first place. According to Google and Amazon, cloud databases will not and do not look like the popular relational databases that are the mainstay of all enterprise applications today. Other players, such as Microsoft are starting with a simplified key-value schema for its cloud computing database, but have indicated that this will keep evolving, presumably to something closer to a full relational model. Finally we have Oracle, which at least until of late has loudly questioned whether the cloud is really anything distinctive at all – which leads me to believe that the company might not have much of a plan yet for a cloud database.
What is clear is this: whether we’re talking about OLTP or Data Warehouses, there will be a pressing need for a variety of database virtualization schemes to support workloads in the cloud. And, the industry has to decide what “database” means when it comes to the cloud. At xkoto, we’ll continue to manage active/active pools of relational databases as a single virtual database, whether those databases are on premise, in the cloud, or straddling the two. It is doubtful that businesses will rewrite critical enterprise applications and abandon their relational schemas in order to take advantage of the cloud. It will be up to software vendors to make sure businesses are able to take full advantage of the cloud.
