Thursday, April 23, 2009

Single-Stack IT: a pipe dream

Overtures of this kind are heard a lot these days:

"Flemmer added that this is a challenge for forward-thinking IT managers, who want to use Microsoft server products but face internal challenges due to the perception that the Microsoft server stack cannot match the performance, stability and security of Unix and Linux-based solutions. “That’s unfortunate because Microsoft has made many strides in the server market in the last few years and it behooves all organizations to look at Microsoft as a viable back-end solution,” he said.

Well, this article -- very much pro-Microsoft, just notice those "forward-thinking IT managers" -- highlights what is in fact, in my view, the key to the successful marketing strategy: sell it to managers, not to engineers. Competing on technical merit can only go so far, as IT history shows. Convince the manager he is forward-thinking, give him a lot of glossy paper to justify the choice, and the technical merit and quality will follow... in many years. Perhaps.

Enter "single-stack" manager, thinking forward to a hefty bonus and a promotion. In a generation, nobody knows how to choose tools to solve a problem, how to program without an IDE, and do a decent comparison study: all that is required is to look it up in a single vendor's catalogue. That this strategy works is shown by the apparent success of the campaign aiming to prove that buying expensive server software is somehow cheaper then getting it free with comparable support.

But this is not the most interesting point in the whole debate, IMHO. What is interesting in that whole debate about "single-stack" is that the whole thing is, in my view, totally misdirected. A single, all-encompassing solution to all IT problems is a pipe dream. What is _not_ a dream is a single-vendor solution, whereby a company becomes dependent on a single vendor for a solution of comparable complexity.

First of all, this has been tried before; the history of IT is a history of blunders and blind alleys (something that makes one laugh at expressions like "industry best practices"; "client-server", anyone?). IBM and, on a smaller scale, Sun, have tried to pull the same trick -- make their customers only theirs. Except that Microsoft seems to have better direction and marketing that would possibly get it even further up the wrong tree.

Then, what is it we are trying to achieve when solving an (inherently complex) problem with a single vendor's set of solutions? What we get is about the same amount of moving parts and entities. All of them need not so slightly different skill sets (look at those job ads, looking for people with experience in different Microsoft products). So we are spending time and money to change the existing mix of products for another mix. Is it simpler? Not that much, technically; but the manager has to speak to a single salesperson backed by a big company with a lot of manager-friendly marketing material.

And, finally, what will happen when we will have acquired the holy Grail? Right. Migrating to the new generation of products from the same vendor. Think of Vista migration, but we will be single-stack by that time, and will have to pony up for _everything_, not just those workstations. Which, incidentally, is the goal of the exercise.

Not that I can't see the advantage of doing a simple thing on a single platform. It's making this a universal recipe that I think is way overboard.

No comments: