WTF is... scale-in?
Big Blue contemplates its server navel
I don't get it when IBM talks about "scale-in" for its new PureSystems converged server platforms. And I bet you don't either, and got the same chuckle at it that the rest of us hardware motorheads here out on the Intertubes did.
Scalability has been an issue as long as there have been systems, and back in the day when a system meant a single machine, you really did need to wonder about how far a machine could scale its performance and memory capacity before you plunked down the company's money and your reputation. Nobody called this vertical scaling or scale up, as they do now, and most times they wanted to know the performance ceiling of a line of machines was far enough above their heads that their applications didn't hit it.
Definitely in, but no scales
In IBM's case, and the advent of the System/38 minicomputer and its integrated relational database management system in 1979 was a breakthrough machine, and one that it took Big Blue nearly a decade to extend with higher performance machines, called the AS/400. The CMOS-based mainframes from the early 1990s also had scale-up issues, and IBM was punished economically for it.
With the increasing use of distributed computing in the 1990s, workloads were broken up into pieces and run on multiple machines – you put the database here, some application servers there, and some Web servers over there, and glue them all together on a network.
Now, you could scale up the database, application, and presentation layers of an n-tier application independently – what is often called horizontal scaling or scale-out in the systems lingo - and you might even use scale-up servers in this scale-out architecture, particularly for the database layer, and that was sometimes called diagonal scaling just to be funny.
So check this funny bit out from IBM's PureSystem announcement earlier this week:
"Scale-In" System Design: With PureSystems, IBM is introducing a new concept in system design that integrates the server, storage, and networking into a highly automated, simple-to-manage machine. Scale-in design provides for increased density – PureSystems can handle twice as many applications compared to some IBM systems, doubling the computing power per square foot of data center space.
I've been in the in crowd a few brief times in my life, and I have an innie belly button like most of you do. My forebears founded the inn in Hartford, Connecticut after being tossed out of the Massachusetts Bay Colony. But I have no idea what IBM is talking about with "scale-in" system design. The more oomph you add, the less space it takes and the smaller your bill gets? That's certainly not it. ®