Last time I looked.... h-store and voltdb have not taken over the database market :-)

But in reality, my view is pretty simple - people like mature, functionally complete, requirement-based fit, cost effective technology.

For analytics, column based stores are great - its a reason we have HCC and in-memory, and other vendors have similar offerings. But do I want (solely) a column based store for high volume transactions ?... probably not. That's a key differentiator for our in-memory product - you get column based access *without* having to run a second database alongside your oltp system.

But I do disagree with the content in two areas

1) that customers need exclusively ram based systems for transactions. A stock standard oracle database (and I'm not being biased here, other vendors I'm sure can do this too) can crank out tens of thousands of transactions per second on commodity hardware. If you need *millions* of transactions per second, then fine ... look at niche solutions, because guess what...you're a niche customer. I'd contend then 99.99% of database customers in the world do *not* need that scale.

2) The concept of "this database is for oltp", "this database is for olap" etc... I think that is rubbish nowadays. Every database is a transactional database, every database is an analytic database - the line is so blurred between them nowadays. Once again, I think that's why the in-memory option is such a cool piece of tech for us.