A question of utility

Application vendors and service providers really, really want your businessâ€"literally.

They've been saying as much for the better part of a decade, talking in long-winded tones about the potential benefits that could flow when a company offloads the complexity of many of its core IT functions to a third-party service provider.

Yet despite their enthusiasm, vendors have so far failed to make much of an impression on their would-be customers, who are largely continuing to run their applications the old-fashioned way. This has been a source of constant frustration for suppliers whose flagging margins could use any guarantees of long-term revenuesâ€"just look at the string of once-optimistic application service providers (ASPs) who have closed their doors due to lack of demand.

Their failure to revolutionise application delivery notwithstanding, many of the industry's leading companies are now spruiking a new line, hoping to reinvigorate an IT market that's been anything but vibrant in recent years.

If you believe the hype, utility computing will do for IT what national power grids did for electricity by making access to corporate computing resources a no-brainer. This will, we're told, happen by effectively building a fifth key public utilityâ€"to join gas, water, electricity, and sewerageâ€"using massive, interconnected arrays of servers. Instead of running applications for themselves, customers will simply tap into the services they need, when they need them, and pay the bill at the end of the month. Sound easy? It's certainly being sold that way.

The utility land rush
Indications are that customers do like the conceptâ€"in broad terms, at least. In one recent survey of 34 corporate IT users, analyst firm IDC found that nearly 65 percent of those customers said they would be interested in utility computing. Recognising the industry's nascent state, IDC advised potential market players to -seize the first-mover opportunity" to establish a position in the market. The market has responded with the usual dose of hype: utility computing has become a rallying cry for IBM, HP, and Sun Microsystems, each of which is working furiously to cement its long-term position within this new and evolving market.

Vendors have so far failed to make much of an impression on their would-be customers, who are largely continuing to run their applications the old-fashioned way.

Yet each company's vision is deeply rooted in its own product strategy: IBM, with its eBusiness on-Demand vision, sees utility computing built around its WebSphere application servers, logically partitionable Linux mainframe servers capable of running thousands of virtual servers at once, self-healing systems and applications, and J2EE-based Web services to link everything together across disparate networks.

HP has focused more on utility computing as a method for accessing additional computing resources, with its Adaptive Networking paradigm envisioning a self-managing, always-available network that provides applications on tap. HP already provides capacity on demand in servers such as its Superdome, which can ship with excess unused processors or storage that can be readily activated by customersâ€"for a feeâ€"as they're needed.

-It's becoming possible to treat software as a service, and to move towards this idea of a business utility," says Roy Brady, strategy and marketing manager with HP Services Asia-Pacific. -A good way to simplify is to standardise a set of design principles around standardisation, consolidation, modularity, and integration. You can then create modularity where you can change parts of your application infrastructure without changing others. It's evolutionary, and builds on a lot of things we've been doing over the last few years."

Sun is taking a similar approach with its N1 architecture, which will aim to build a self-managing utility computing ecosystem that leverages the rich heritage of the popular Solaris operating system and Sun's ongoing commitment to the Java it created.

Yet early estimates about these architectures' takeup have proven optimistic at best: Sun, for one, initially envisaged the introduction of N1 as a three-part process, with customers stepping through virtualisation, services provisioning, and policy automation phases in 2002, 2003, and 2004 respectively. Now, halfway through that timeline, most companies are still only vaguely aware of the concept of utility computing, and very few have taken any steps towards anything that could be termed utility computing.

Those that have adopted something like it have done so as part of larger outsourcing deals. Companies like JP Morgan Chase, AXA Insurance, Deutsche Bank, and Canada Life all recently entered into multi-billion-dollar outsourcing deals with IBM Global Services that involve large-scale server consolidation and the ability to access extra capacity when it's needed.

But those deals aren't really about building a computing utility as much as they are about natural centralisation; there is no revolutionary method of application delivery, no new technology to be used here. IDC saw these companies' conservative approach to utility computing as a way of minimising the risk of what is inherently a very risky change. If companies committed to the utility computing model early on, they could easily not only lose control of their IT, but be left without a backup if something went wrong.

If it works like an ASP, and sounds like an ASP...
These concerns may well sound familiar: cast your mind back a few years, and you'll recall a similarly disruptive philosophy that promised to change the way applications were delivered. Application Service Providers (ASPs), which stormed the markets in 1998 with promises of slashing implementation times by allowing customers to rent access to hosted enterprise applications, once seemed set to change the world.

Yet those companiesâ€"many fuelled by billions in investment reaped from a still-gullible dot-com investment communityâ€"were forgetting that reality is a harsh mistress. Although it was technically possible for companies to run major applications like SAP R/3 on another company's computers and rent access to that application, it soon became clear that commercial ASP services were not sustainable at the kind of scale needed for a viable business model.

Well-funded US companies like Pandesic, HotOffice, and Red Gorilla closed their doors, while Australian companies like Peakhour, ZLand, ASP Pty Ltd, Tequinox, and Managed IT either evaporated or were quietly subsumed back into their corporate parents as managed services divisions. Although ASP services still exist at firms like Allegiance Systemsâ€"an outsourced payroll provider that invested $3.8 million to shift its entire business to ASP several years agoâ€"they've been most successful in providing specific, niche-oriented applications or services that have long been outsourced in the past, such as payroll processing. Other tightly scoped managed servicesâ€"for example, accounting packages, productivity tools, or hosted Microsoft Exchangeâ€"have enjoyed limited success.

Companies like JP Morgan Chase, AXA Insurance, Deutsche Bank, and Canada Life all recently entered into multi-billion-dollar outsourcing deals with IBM Global Services that involve large-scale server consolidation and the ability to access extra capacity when it's needed.

Customers' lukewarm response to the ASP movement highlighted the intrinsic problems with vendors' assumptions that customers were ready to delegate responsibility for their total IT infrastructures to third-party providers. Although providers claimed they could use templates to serve enterprise apps to many customers at once, those customers were generally unwilling to accommodate the sameness that the ASP model provides.

Now that the ASP concept has remerged under the utility computing banner, corporate IT users are facing similar decisions to those they faced several years ago. Vendors are falling over themselves to prove that this time, it will work better.

HP's Brady, for one, likens utility computing's current early days to the early days of electricity, when each city had its own generator and there was -a plethora of different standards". Martin Duursma, vice president of advanced products with Citrix Systems, says utility computing will make applications work -just like a tapâ€"you open it and get water, and the thing is available all the time."

While it may sound cute, there are major problems with this analogy. The utility model vendors are describing devolves the chain of responsibility to the point where data is seen as a generic, public service provided by a faceless corporation. Yet data is not a generic commodity like other utilitiesâ€"after all, water doesn't come out of the tap in different flavours and no two types of data are the same.

This very important issue muddies the argument for utility computing, according to Rajat Kumar, senior advisor for advanced technologies with Microsoft Australia. -Electricity is about how you power things up, and it's fairly standard," he explains.

-If computing was based on everyone owning the same bits, it would be a simple problem. But we have different schemas and application interfacesâ€"all the dependencies which make the utility concept different than computing. If you think of utility computing as being purely cycles oriented, you start to think of it as a utilityâ€"but if you start to ask what you're going to do with these cycles, the metaphor changes and it's not a utility anymore."

New paradigm or marketing play?
Resolving this paradox is a major obstacle for the companies developing the infrastructure for utility computing. As anybody who's ever worked on an IT project knows, today's applications are complex assemblies of thick and thin clients, middleware, centralised and distributed databases, Web and application servers, service monitors, technical support, and resolution services.

For utility computing to provide these capabilities over a generic, public infrastructure will require considerable progress on the part of application developersâ€"if it can be done at all. Yet there is hope: corporate developers have at their disposal many tools that were still on the drawing boards when ASP providers were trying to make their mark.

Most notably, Web services have emerged as a mechanism for joining application components in widely disparate locations. Both .NET and J2EE Web services models will allow applications to work seamlessly across the kind of managed environment that's intrinsic to utility computing. Customers are still in very early days when it comes to Web services, although initial experiments have demonstrated that the concept is viable even as standards bodies continue to develop its vocabulary.

The other major element of utility computing is management, with increasingly intelligent systems being designed to monitor service performance and resource allocation. HP, naturally, is addressing this requirement with its ubiquitous OpenView software, while IBM and Sun have their own management strategies. -It's much more sophisticated and solid than ASP because the vendors themselves are building their applications as components and as a service-oriented architecture," says HP's Brady.

Recurring roles for existing products confirm that early utility computing models, at least, will largely be assemblages of existing technologies, modified slightly to run with as little intervention as possible. Importantly, these tools are evolving in a way such that computing utilities may in fact be developed in-house and used to deliver services to business customers internally.

From that perspective, utility computing becomes little more than a branding exercise, a new way of referring to technologies that are already in place. Those technologies are being rapidly improved with the kind of self-awareness necessary to maintain utility-like availability levels. IBM, for one, has been particularly vigilant in outfitting its applications with self-monitoring and self-repairing capabilities that would be essential for utility computing to work as its advocates describe.

-This means that people can get away from having to worry about local conditions in terms of how they consume those services," says Citrix's Duursma, who believes the architectural components of utility computing are largely in place but that service management, usage metering, and the ability to charge customers for discrete services still need improvement. -All the pieces are there, but all of themâ€"including those from Citrixâ€"can be simplified."

The new utility grid
From an application perspective, utility computing is primarily about distributing application components across a highly-available network. In most discussions so far, details of that network have been somewhat hazy, with the Internet or some private IP network the assumed carriage mechanisms. But in the kind of model that vendors are describing, the networkâ€"an always-on, highly available infrastructure that's internally redundant and universally available in the same way that the electricity grid is universally availableâ€"will be just as important.

Efforts to address the network side of utility computing got a symbolic if not earth-shattering shot in the arm last month, when researchers from the Universities of Sydney, Melbourne, Adelaide, and Australian National University demonstrated a prototype computing -grid" that used software to distribute a massive number-crunching exercise between Linux-based desktop PCs at each site.

Using locally-developed GridBus software and the Globus Toolkit (www.globus.org) grid middleware, researchers were able to speed analysis of a 10TB data set by using the grid controlling software to distribute data between the systems. Results were then delivered to the researchers, who could benefit from the analysis without each having to have a copy of the data.

Raj Buyya, head of Melbourne University's Grid Computing and Distributed Systems Laboratory, says the successful trial demonstrates the value that utility computing could play for businesses requiring complex analytical capabilities only part of the time.

-A business might need a huge-precision capability just once a month to do sales reports," he says. -The grid is a key enabler for this kind of computing, by allowing [companies] to consume these capabilities based on their quality of service requirements. If a service is very urgent, they may be happy to pay a higher price for it. But you need computer technology that understands this."

That prototype ran over the Internet, although future computing grids will likely be set up on top of high-speed infrastructure like the AARNet-led GrangeNet or CSIRO-directed CeNTIE grid computing networks. If utility computing takes off over time, telecommunications carriers will likely set up their own dedicated fibre networks to support high performance and availability.

Ultimately, utility computing grid networks will require a common management core so that the architectural differences between nodes can be de-emphasised. Web services provide this level of abstraction at the application level, while Globus is an example of a lower-level grid controller that will be necessary to keep grid nodes synchronised. Using such technology will allow utility computing providers to distribute the applications they offer in a more effective and efficient manner.

The Globus Project is well on the way to setting the de facto standard for utility computing, a trend that was boosted when IBM recently said it would build Globus' Web services-based Open Grid Services Architecture (OGSA) into Linux, AIX, eSeries mainframe servers, WebSphere application servers, and its DB2 database. The company's wholesale commitment to grid computing will set the pace for other competitors, potentially jump-starting applications' awareness of utility computing infrastructure.

Where to from here?
Customers are inherently sceptical of any major architectural change, and it will be up to would-be providers to convince them that the model is viable and worth pursuing. Over the next few years, businesses can expect vendors to become increasingly aggressive about utility computing and its benefits to their organisations. By 2006, Gartner Dataquest recently predicted that customers will have warmed to the concept to the point where subscriptions to utility services make up 31 percent of the IT management services market.

Customers aren't the only ones that will have to change to accommodate utility computing, however. It will also force integrators, ISVs, and other members of the IT food chain to reassess their roles and reshape their services accordingly. Because utility computing can potentially disenfranchise many intermediaries currently involved in IT service delivery, analysts are warning that service providers will have to reassess their value to customers and change to adapt the utility model.

-Few outsourcing vendors will have the ability or resources to continue adding value at the high value-added/high-cost end of the market," the firm predicts, suggesting that many more outsourcing vendors will have to operate -in the lower value-added, commoditised arena where the critical success factor will be cost consciousness. A far-from-glorious fate awaits many vendors that are in denial as the true consequences of customers' growing embrace of lower-cost utility computing become apparent." Nearly 25 percent of traditional companies will, Gartner Dataquest predicts, fail to adapt and will simply disappear.

Such claims promise significant disruption to the IT market as utility computing becomes entrenchedâ€"but it's not the first time analysts and vendors have gotten whipped into a frenzy in their rush to correctly assess the market opportunities presented by the new technology. Customers may find the promise of utility computing tantalising, but they need to be realistic about its rollout. The fact that it is very much a work in progress means it will be years before those companies should consider doing anything more than educating themselves.

EXECUTIVE SUMMARY
		The idea of the computing utility is driving much of the industry's product and strategy development, but we're still a long way before this technology will work like it's been described. Here are a few things to think about before you get into it.  •  This is all fantasy...for now, at least. Everyone has more or less agreed how utility computing will workâ€"conceptually, at least. But plans change rapidly, and unknown forces may affect the technology's final shape in ways that are hard to predict now.  •  ...but vendors will make it real. IT suppliers, who prefer predictable and ongoing revenue streams to the uncertainty of the last two years, will make utility computing happen. Expect emerging interoperability as Web services evolve and new standards, such as the OGSA technology IBM will build into its key products next year, clarify the computing utility's terms of engagement.  •  Know your supplier. Right now, the term -utility computing" is largely a branding exercise, with IBM, HP, and Sun all building their utility strategies around combinations of existing products. Many other companies are developing their own plans, however, so it's important to check with key suppliers to see where they stand.  •  Virtualisation is the key. Whatever technical complexity is hidden in the background, utility computing is all about virtualisation--the wanton creation and destruction of logical servers, storage space and other resources that can be allocated to different customers at will.  •  Grids aren't there yet. Grid computing--spreading compute loads across myriad servers located in different places--is a core enabler of utility computing. But grid technology is still in its infancy, with standards being set and network owners conducting their first tentative trials now. Don't depend on commercial grids appearing any time soon.  •  A utility can be internal, too. Although the grand scheme for utility computing involves the availability of services over a ubiquitous shared network, you may find the technologies equally useful--and more readily accessible--if they're run over your internal network. Consider what benefits might flow from turning your internal IT organisation into a utility provider.  •  This is ASP... but then again it's not. The disastrous application service provider (ASP) movement is the closest analogy to utility computing, which doesn't bode well for the new paradigm's advocate. But ASPs bit off more than they could chew by focusing on high-end enterprise apps; the proper role of an ASP, and the most likely early target for a computing utility, is to provide technology enabling specific business functions in bite-sized chunks.  •  Think modular. As with any type of outsourcing arrangement, utility computing requires businesses to carefully consider their current and future application needs, then identify synergies, redundancies and potential areas of improvement.  •  Utility computing is risky. There are no two ways about it: what vendors are proposing is difficult technically, particularly since it requires a far better understanding of issues like security, detailed usage billing, and the like. Furthermore, if it stops working for any reason, you won't have a backup. Ignore analysts' heady predictions, and take it very, very slow before you jump into utility computing.