Blogs

RedMonk

Skip to content

System z Software Summit 2006: Mainframes and the Distributed Developer


One of the major topics I’m typically interested in is figuring out how companies can sell into the long tail of the tech-market. With the dot com crash and the following consolidations and burn outs, the fat-head (Fortune SomeNumberHere companies) is mostly gobbled up by the major vendors. More importantly, battling over the fat-head seems like a closed loop: while you might win over a giant account on one front from your enemy vendor, your enemy vendor is probably going to win an account from you that balances that gain.

The even more germane fact is that if even a small percentage of your fat-head customers burn out — say American airlines, auto-manufactorers, or telcos — your revenues aren’t going to look so good for several quarters.

Finding New Mainframe Customers

From that comes our emphasis on lowering the barriers to entry: the micro and small buyers in the long tail don’t have the time or money to spend on learning or purchasing traditional enterprise software and hardware. While mainframe sales account for “about half” of IBM’s profit, that base is steadily shrinking. More importantly, IBM has effectively maxed out the current mainframe market: there’s not many more sales in the current customer base.

Those two things mean one thing: it’s time to get new customers, new types of customers.

The New Mainframe Customer

IBM needs to start selling more mainframes to “distributed” customers. That word — “distributed” — points to the primary problem of accomplish that task: culture. For those who haven’t been in the mainframe world, “distributed” is mainframe-culture talk for anything that isn’t mainframe oriented. The even more telling part is, distributed folks usually don’t even know the label the mainframe people give them.

Here’s a funny story that illustrates the schism:

While still at BMC, I was getting a coffee with the usual suspects. I’m not sure what we were talking about, but it was probably a mix of The Big Lebowski, blowing off steam about adopting Scrum, or some Java. An older woman behind us was patiently waiting in line for coffee, and as we were leaving the break room, she said, “are ya’ll distributed people?”

“Yes, we work on BPM,” one of us said.

“I thought so. You talk like distributed people.”

And that seems to just about wrap up the challenge of selling mainframes to new customers.

At least we have coffee in common.

diff mainframe distributed

The two cultures are in completely different silos when it comes to (a.) spending habits, (b.) technology, and (c.) development practices. I don’t have the stats on hand to back it up, but it’s common gut-knowledge that age is always a major difference between the distributed and mainframe camps.

IBM’s z9 Business class is targeting the pricing of mainframes, the introduction of Linux and Java on the mainframe is a great way of addressing the technology differences, but I’m not sure the core culture problems — the development practices — are fully tackled.

White-boxes Are Cheap

At the IBM System z Software Summit, the pitch I heard again and again was that mainframes have great Total Cost of Ownership, or TCO. To be a cynic, what that means is that you’re going to pay a lot up front, and even in the middle, but eventually it’ll pay off. That pitch is completely the opposite of how distributed, long-tail developers think:
I’d gladly pay more on Tuesday for rapid development today.

From the perspective of a mainframe mindset, distributed developers are cowboys: we’d rather get some crap boxes and get something up and running as quickly as possible rather than take the time to study what hardware we actually need. This is why and how companies like Yahoo! and Google ended up with server farms and the custom OS coding and sysadmins needed to maintain them. For all the arguing that the up and comers do about “all you need are a bunch of white-boxes,” I often wonder if that’s more the case of the budget the companies had at the time rather than the technological facts. That is, would Google work on 10-100 mainframes vs. 1,000’s of white-boxes?

Google on a Mainframe

As a side-note, that question itself is a good one for IBM: would mainframes be a better option for Google, meaning, would Google work better and be cheaper to run with mainframes? If the answer is no, then there’s probably some re-jiggering to do.

The reason is because Google’s white-box server farm is the understanding of hardware planning and requirements for the long-tail developer: do it like Google, and you’ll be successful. And until/if Google changes how it runs it’s IT or tanks to below $20/share, there’s no arguing against Google’s IT practices in the minds of the long-tail developers.

Mainframe Exposure

The primary pieces of advice to get more exposure for the mainframe in the “distributed” world are:

  • Ship it with a fully functioning LAMP stack to appeal to the LAMP crowd. I don’t know for a fact, but getting LAMP to work on a z is probably a day or two of hassle…not to mention figuring out the mainframe licensing.
  • Take an actual z9 to all the long-tail developer [un]conferences. It’s a rare thing for a developer to actually see a mainframe, let alone touch it. The positive effects of being a few feet away from a mainframe may seem cheesy, but they’re real: all dorks are driven by gadgetfetishes, and there’s nothing wrong with feeding that.
  • Have code-fests on mainframes at these conferences and unconferences, by grabbing some huge datasets and encouraging developers to do mainframe mashups.
  • Most importantly, introduce pricing or access that allows long-tail developers to use mainframes without fussing about how to pay for them until they have the time to do so.

Now, many of the “distributed” developers out there will happily ignore the above, and go on using white- and black-boxes. That’s fine. What’s important is taking a distributed mind-set and getting out into the long-tail as quickly as possible, neverminding the perfection of that journey. Waterfall vs. Agile. It’s either that or eat off the fat-head for years to come.

Disclaimer: IBM, Sun, and BMC are clients.

Technorati Tags: , , , , , , , , ,

Categories: Community, Compliance, Conferences, Marketing, Programming.