RealTime IT News
What's Wrong With On Demand?
By Erin Joyce
May 27, 2005

Computing On Demand. Sounds so straightforward, doesn't it?

Computing services on demand, by the drink, as you need it, utility style. Makes sense to the bean counters.

As IBM describes it, the term means offering data center services like a utility, and basing those services on open standards that underpin integrated systems imbued with self-healing (autonomic) capabilities.

So why isn't the term catching on with IT buyers at a better clip? Because On Demand describes too many functions at once. And when you try to pitch a bean counter on too many ideas at once -- or a concept that touches just about every layer of the stack -- all of a sudden you're walking in the land of the fuzzy, vague and sometimes wishy-washy business ideas.

Wishy-washy won't get a bean counter to say yay or nay, or look at the bottom line. And that could explain why, for all the cost-savings computing On Demand could mean, many companies are just not buying into the concept. At least not now.

JP Morgan Chase is the current best example. In 2002, the bank awarded IBM a $5 billion outsourcing contract, with On Demand services built in. At the time, the deal was seen as the benchmark for On Demand deals. Big Blue was supposed to take on a significant portion of the bank's data processing infrastructure, including hosting data centers, help desks, distributing applications, and maintaining data and voice networks.

Maybe it was this part of the deal that gave JP Morgan pause: IBM was also expected to help power and host mission-critical functions such as trading applications for the securities side of the banking giant's operations. There were more promises too, such as creating "virtual pools" of computing -- an early term for grid computing services.

For plenty of reasons, the bank scratched. Granted, blending its platforms with merger partner BankOne, which came after the contract was signed, is a much larger priority right now. But guys who work in the back end of the bank told me they had cold feet about offloading that big a chunk of their computing systems to a vendor, only to have it served up to them like a utility.

This is the reality for systems vendors right now. It's taking longer for internal cultures, as well as senior management making buying decisions, to warm to the approach -- and to the depth of work it will take to achieve this kind of computing anytime you want it.

If you're a software vendor, you're probably having an easier time with On Demand. Software by the seat provider Salesforce.com is the best example of a company providing customer relationship software to customers by the drink, which they can then customize themselves. Software as a service, software by subscription. Why not? It's much easier in the software industry, especially as more people discover that much of what they do on the Web each day is, in part, a kind of on-demand software service.

But don't think all is rosy in Software On Demand land either. In one On Demand model, customers get the software delivered to them by the network. They play around with easy pull-down menus, and customize it without having to interface with prickly programmers. Sounds great. Until the vendor delivers an upgrade.

It can become chaotic maintaining old APIs and new ones, especially for smaller customers. Do they want to be in the business of programming themselves? All of a sudden, the lovely idea of On Demand software -- having somebody else deal with the development and maintenance costs -- is back on the customer's plate again.

That's the downside to the shift away from burning software on a CD and shipping it to customers to configure as they will. Already, industry analysts are racing to release research reports warning about the bewildering array of APIs and difficulty that On Demand providers such as Salesforce.com are running into when it's time to deliver software upgrades On Demand.

There's another problem. Because the concept is so wishy-washy and applied to both software-as-service and the huge infrastructure plays promoted by IBM, customers aren't always sure exactly what they're being pitched.

Bruce Cleveland, senior vice president of On Demand for small businesses at Siebel, says one reason for the umbrella term is because we're at a relatively early point for shifting to this kind of technology delivery.

"These kind of labels being attached to these services go beyond what you would consider to be application managed services, where traditionally one would outsource computing to an EDS, who would then take care of the software management," he says.

That said, he believes the industry hasn't done as good a job as it could in distinguishing the different versions of On Demand computing that are out there -- not to mention the different price models.

But make no mistake: On Demand is happening. And eventually, the baby-step pace will become a jog, then a sprint, then a race.

John Patrick, an industry visionary who helped craft IBM's Internet and On Demand strategy -- among many other things during his more than 30 years with the company -- says explaining On Demand is really about explaining an attitude.

"It says that all of the data and all of the processes that are needed by any of the constituents of that organization will be available to them whenever they need it, wherever they are -- with any device they might be using to connect to the Internet," he tells internetnews.com. "So it's a simple folksy way of saying, whenever you need it, you can get it."

Sounds great. But simple on the front end means not so simple on the back end. There's the big picture idea, then the need to bring in the details people to drill into how to make it happen.

That's why moving to a true On Demand computing system means talking about virtualization, hosting applications, resilient infrastructure, failover databases (after all, this is On Demand we're talking about), and other systems management concepts.

This isn't to say companies don't get it. They're just dipping one toe in at a time, while they figure out how to re-architecture 1960's era mainframes and get the moats around data silos filled in so different databases can talk to each other. This helps explain why IBM's WebSphere line of middleware continues to sell at a brisk pace.

Siebel's Cleveland says the story of system vendor Digital Equipment's demise is a good way to explain why software and computing On Demand are inevitable --no matter how slow or confusing the process right now.

Another era ago, the venerable DEC built a fantastic operating system, among other things. It had thousands of engineers building hardware for it and working on proprietary features like clustering to make them appear as an integrated whole. It was all quite brilliant.

Then the Unix operating system started to evolve, as then-owner AT&T began making modifications to the base operating system on behalf of all the other competing hardware providers.

"All of these companies shared the cost of development of the operating system," he says. "They were able to create a product offering that ran the same application, same database, same system management software that DEC ran -- at substantially reduced costs."

Once large companies realized they could buy the same computing power at 10 times less the cost, the handwriting was on the wall for DEC.

Many schools began teaching Unix, and droves of Unix-savvy engineers were soon swarming the industry. DEC was doomed.

There's more to DEC's demise, of course. But in some ways, a similar tale is unfolding in the software industry, as well as among systems vendors, as companies realize that internal costs of developing and maintaining their computing systems can be shifted to those providing the computing service.

Providers are well aware of DEC's cautionary tale, especially how its faltering entry into workstations laid low the once-mighty company. They realize another shift is underway and better stay afoot, and at the right pace.

[On Demand] "starts with this attitude that it can, and will, be done," says Patrick, who happens to be the head of a consulting firm called Attitude LLC. (Full disclosure: Patrick also sits on the board of directors of Jupitermedia, the parent company of this publication.)

The day one airline is able to let people switch flights, or check travel data from their smart phone, or make secure data calls to the enterprise via Web services -- and see how wild their customers are about it -- you will see and hear a thundering herd of companies signing up for the work of delivering computing and data, On Demand, by the drink, utility style.

By Patrick's reckoning, we're at about 5 percent uptake of On Demand services. There's a long way to go yet.

Erin Joyce is executive editor of internet.com's news channel