Original URL: http://www.theregister.co.uk/2006/12/21/utility_computing/

Time to denationalise the software monopolies

From utility to commodity

By Gavin Clarke

Posted in Applications, 21st December 2006 10:54 GMT

Analysis As PR stunts go, Sun Microsystems' Blackbox trailer park computing launch this autumn wasn't bad. Sun pumped up some rather dull AMD Sun servers by unveiling a black, spray-painted, branded shipping container that could be crammed with Sun storage and server devices.

Sun director of web technologies Tim Bray summed up the spirit of the event, saying it wasn't about the identifiable market, but the concept: "I'm glad we built it, because it is just totally drop-dead fucking cool." While people debated where you'd use such a device – the corporate car park or an oil delta - the trick worked and everyone from bloggers to trade publications and the national press covered trailer park computing.

For Sun, it was the epitome of chief executive Jonathan Schwartz's dream for utility computing. The definition of utility is questionable, but utilities are - generally - boring things that are invisible to the end-user and become available to the consumer at the flip of a switch.

Utility computing is Silicon Valley's latest uber trend. Beneath the talk of software as a service (SaaS), Linux on 32-bit servers, and infrastructure management is a belief that powerful and productive systems can be delivered, and charged, as if they are electricity or the telephone service.

While SaaS and Linux are certainly changing the landscape, they are the latest manifestation of a trend that has been snaking its way through computing for years: to make IT a more standardided, repeatable and a more controllable process.

Utility, though, was realised at an infrastructure level years ago, with the rollout of PC and server software and middleware. That means modern-day, utility based computing is already here. The goal now is liberating that infrastructure from the control of the few who built it for the benefit of business users, by turning it into a genuine commodity that is open to competition on pricing and supplier choice.

As with the denationalisation of electricity and telephone monopolies in Europe and the US, though, the transition to commodity faces an uphill battle that will see incumbents try to crush the start-ups using politics and economies of scale.

Business is a rich killing ground for utility and commodity computing. They bring consistency, predictability and choice to products and pricing, and end single vendor lock in.

The market certainly needs it. Big vendors have built their business selling software licenses and hardware that customers don't really need and can't really use. It was irony, indeed, when Microsoft chief executive Steve Ballmer joked with delegates at a Microsoft conference he didn't – or couldn't - use all the features in the Siebel customer relationship management (CRM) software that Microsoft uses.

Talk about coals to Newcastle. Siebel with other CRM vendors got a bad name during the 1990s for being difficult to install and use, and for stuffing software licenses down the customers' throats. The birth of Salesforce.com can be traced back to this dark time. Microsoft, meanwhile, is renowned for confusing and confounding users using a matrix of server, client, client access, and volume licenses.

Companies like Microsoft laid the foundation of today's computing infrastructure by providing software on desktops and servers that performs an important role but that is, ultimately, easy to overlook and boring to the outside observer. It's only when this software is not around that it becomes interesting because you miss it, like power or the phone going down.

Take Office. Word, Excel, PowerPoint and Outlook automated what humans did with paper, pen and the telephone. Most users employ Office literally for the basics – to write word documents, make spreadsheets, and send emails - that's why upgrades are so slow. Office is a functioning utility. It also runs on 90 per cent of the world's PCs, meaning it's achieved the level of presence demanded to be considered a genuine utility.

Middleware can also be considered a utility. It covers a set of products that – like electricity and the phone line – are invisible to the end-user but keep things ticking and are missed when they are gone. Middleware sits in the infrastructure layer of large organisations' IT performing important - but mundane and repeatable - processes like serving up data, forwarding traffic, and policing user access.

Despite their status as utilities, Office and middleware are not charged as utilities or treated as commodities. Microsoft will charge $400 per copy of Office 2007 – a price many consider expensive and will ensure a good percentage of users do not switch over, overnight.

On middleware, Oracle's enterprise database, for example, kicks in at $40,000 per CPU. Middleware rival BEA Systems charges $10,000 per CPU for WebLogic Server, despite the consensus that the application server is the weakest of all middleware's submarkets. Microsoft, Oracle, BEA and others spent decades developing and building these software infrastructures, so – clearly – feel they have the right to charge accordingly. Often this is justified in terms of returning value to their shareholders. Like the bad old days of telecoms, too, switching is difficult: businesses can't easily change middleware supplier because of vendor lock in and the enormity of the change process.

What we are seeing now, though, with Linux and SaaS vendors is a challenge to the status quo. Companies are either capitalising on the underlying infrastructure or offering a more open infrastructure that is relatively easy to change and is more affordable. They are commodity companies. And that's why the utility providers like Microsoft and Oracle are falling back on old tricks and vendor politics to maintain their monopoly.

The computer industry has been toying with ideas of utility and commodity for a while. Last decade, Unified Modelling Language and Common Object Request Broker (CORBA) from the Object Management Group (OMG) were the vision for utility and commodificaiton in software. Both promised to revolutionise the way software was built and delivered by introducing a standardised language for building software and enabling the re-use of software components - as if they were pieces of a machine. In other words: software as utility, interchangeable and – ultimately priced – as a commodity.

Their downfall was technological and political. Despite the promises, UML and CORBA were too complex for the mass market, meaning they could not achieve the ubiquity of utilities.

They were further compromised by vendor politics. Patchy and half hearted backing came from companies that wanted to ensure customers continued to use their own products rather than switch. Microsoft put a final nail in UML's coffin last year when it launched its own solution to the design and architecting of applications with Visual Studio Team System, in which UML is just one approach among many others for building applications. CORBA, meanwhile, was overtaken by Java, which became the flag to which non-Windows, COM, DCOM and .NET vendors rallied.

While Java and .NET became the building blocks for today's middleware utilities, the idea of commodity could not be realised. Vendors began adding what they called "value" on top of the basic infrastructure, with features for scalability, performance, security, integration and simplification, which they use to justify high prices.

Now, in the 2000s, the big vendors are mobilising again to contain the threat posed to their monopoly on the utility by the slide into commodity promised by SaaS and open source.

Faced with the threat of another BEA on application servers, this time from open source, IBM bought Gluecode in 2005. In late 2006, Oracle decided it would undercut Red Hat, and Microsoft has partnered with open source application and middleware providers to optimise Windows to their software at the developer and runtime level.

What we are seeing from those who built out the software computing utility is an attempt to embrace commodity using only the elements that suit them. Far from undergoing some Saul-like conversion, the clever ones are either trying to squash rivals or buy them, to contain the threat while maintaining their existing monopoly.

We've seen plenty of hype around utility computing. The utility, though, arrived years ago with middleware and applications that have become ubiquitous and invisible to the end user, like electricity or dial tone. The real story is the transition to price competition and supplier choice that accompany commodity. Like European and US telecoms and power monopolies in the 1980s and 1990s, those who laid this utility infrastructure will be the ones who resist change using elaborate tactics to maintain high prices and lock in. ®