September 20, 2010 3 Comments
Exacloud, CloudLogic, Exalogic aka “Cloud in a box” sounds to me like you’re buying way more capacity than you need.
I like the idea of buying these big boxes with everything pre-configured appropriately.
After all, I’m waiting to embark on an upgrade to 11gR2 but what’s holding it up is that the boxes need to be built and configured internally which takes lots of time and money and red tape. And probably won’t be done 100% right first time…
But on the other hand, aren’t these big boxes all a bit back to the future?
It also strikes me that infrastructure is on a completely different trend at the minute compared to modern, agile software development trends with YAGNI principles etc.
If you don’t really, really need it now, why pay today’s price for tomorrow’s infrastructure?
But hey, what do I know about middleware apart from what it tends to do to my databases…
Finally, a quote from Marcel Kratochvil’s thought-provocative Foreign Key discussion paper:
As for Middleware, I will finish with the observation that most sites do not require Middleware. Scalability can be achieved by efficient database programming and ensuring the database logic and objects reside next to each other. Middleware only offers solutions for sites that require serious scalability or have legacy systems. Generally sites using a Middleware layer will produce a culture that encourages inefficient programming habits because they insist on hiding the database from the programmer (who might not even know what the database is) and adopt the mentality that the bandwidth between the middle tier and the database is infinite. This will not lead to a scalable application but rather the opposite.
Actually, that wasn’t the quote I went looking for. But it seemed to fit the context. What I was looking for was this one from the same source, which is what Exalogic and it’s “direct connection to Exadata” initially reminded me of slightly:
So within five years we will start to see the move away from 3 tier and back to a 2 tier model (but not the client/server model) for most applications. It will prove to be cheaper, easier to manage and will be scalable. The middleware layer will move into the database and the database server will become more sophisticated.