The end of client/server as we know it?

Back in the mid-90s, client/server was the 'paradigm' of the moment, and the PC was at its heart. But it soon became clear that this particular 'paradigm' was expensive and inefficient.

The Gartner Group calculated the true cost of client/server, focusing on support and maintenance of the desktop computer, not just the initial purchase cost of the software and hardware. 'Islands of information' began to rise out of the data quagmire.

But the PC players fought back. To combat escalating costs, they improved their network and desktop management software. This brought down the total cost of ownership, making it easier for IT departments to justify the continued existence of the PC. Data warehousing and other replication and consolidation techniques helped provide the previously missing unified view of corporate affairs.

But then another threat came along in the guise of the network computer (NC). Oracle CEO, Larry Ellison, talked up the concept in 1995, hailing it as a cheaper, more flexible computing 'paradigm'. However, client/server saw off this threat too - again, the PC industry concentrated on improving manageability, and end users proved unwilling to ditch the PC.

Mark Raphael, program director at Meta Group, said: "The NC died an extraordinary death. The reason for this is that Oracle is a database company and would do well to remember that... Oracle is not in a position to set a trend for company architecture."

However, the good old PC - and some would say client/server - are now facing what may prove to be their stiffest challenge, a challenge which, arguably for the first time, is about applications, and not just cost.

Four years after Ellison's original NC speech, Silicon.com spoke to Oracle's UK managing director, Phillip Crawford (the full interview can be seen in the Knowledge Management Channel). He defended the NC concept, but described it as an evolution: what was once a term applied to a specific bit of hardware has become a more generic way of describing a new breed of access devices, whether they be PDAs (personal digital assistants), kiosks, mobile phones or set-top boxes.

Significantly, Web technologies are binding all these devices together. More companies are opening up their data to customers; more are gearing up their mobile workers with laptops and PDAs. Human resources (HR) departments are beginning to employ kiosks to allow staff to update their personnel records.

Increasingly, these applications and the data are accessed via browsers. But this doesn't mean the end of client/server. Browsers may be a new form of client software, but they're a client nevertheless.

So has the Web killed off client/server? No. Meta Group's Raphael believes it's alive and well, and - while its deployment is changing, with many companies looking to consolidate servers for example - it is now being used as a base for that most modern of business models, ecommerce.

Hugh Jenkins, enterprise product marketing manager at Hewlett-Packard, said: "Client/server never killed off the mainframe, so why would the Web kill client/server?"

As far as the industry goes, the vendors see 'Webification' as a straightforward extension of client/sever. Web technology is just an additional layer to the client/server architecture, they claim. Jenkins added: "A decade from now, a lot of systems will become Web-enabled and the same people who deliver the software today will be delivering the software for any future developments."

He continued that companies will indeed migrate their client/server systems to Web technologies. The key advantage is that organisations can scale applications to address the entire population of the world - all you need to access a service is to fire up a browser.

But Rob Hailstone, research director at Bloor Research, hopes client/server does disappear from some areas. The problem with the setup, he claims, is that companies deploy the idea in one place and when it works, decide to move everything over to that style of architecture. What they don't realise is that what's good for one requirement may not be good for others.

With more workers accessing information and applications from out of the office, all that is needed is a device with the ability to access the Web, send emails and hook-up to the office network. There's no reason why IP-based networks, with the bulk of the processing power held centrally and information accessed via a range of devices - including the PC - shouldn't become 'the computing paradigm for the new millennium', as the marketing folk might put it.

That isn't the network computer, but it is network computing. And it isn't 'true' client/server. It's a combination of the last three combatants in the battle for IT directors' mindspace. The mainframe had its faults. Client/server had its faults. The NC hardly got off the ground. But largely thanks to the Web, all of these can co-exist.

Client/server refuses to roll over and die.