A recent post from Mark Perham's blog titled "The Emperor has no Clothes" suggests that distributed database systems may not be as useful as many believe them to be. Perham's opinion, in fact, is very concise:
My belief is simple: avoid distributed databases if possible. You will pay a heavy tax for their use.
However, his argument doesn't come without defense, and his qualification of "if possible" is key. There are scenarios, which Perham lays out, in which it is beneficial or even necessary to use distributed databases, but there are also many in which a more "old-fashioned" database such as SQL may be completely adequate, or even superior.
Is there too much hype surrounding distributed databases? The argument brings to mind the recent criticism in Big Data of a perceived obsession with using Hadoop for everything (every dataset may not call for it), but the reality of network failures is a serious issue, and it is important to take precautions.
What do you think? Does Perham have a point, or is there more to the story when it comes to distributed databases? Leave us a comment and let us know.