Back to the mainframe for the next big thing

According to Oracle, the grid will be the next big thing for computing. Last week, speaking to a packed audience of users,...

According to Oracle, the grid will be the next big thing for computing. Last week, speaking to a packed audience of users, independent software developers and services companies from around the world, Oracle chief Larry Ellison launched what he sees as a milestone product, not only for Oracle, but also for IT users.

Ellison is used to making bold statements that seek to change the direction of the industry. In the early 1990s, along with companies such as Sybase and Informix, Oracle played a key role in selling the benefits of client-server computing.

As it turned out, client-server applications were notoriously ineffective, prone to failure and, with hindsight, damaged the reputation of IT far more than the mainframes they supposedly replaced.

And who can forget the network computer? This big idea saw the industry rally behind a campaign to rid the world of PCs. Oracle even created a subsidiary, NCI, to develop the idea of Java-based network computers as an easy-to-manage alternative to desktop PCs. NCI fell by the wayside, and while network computers are being used as PC alternatives in some applications, they have failed to live up to the hype generated by Ellison and Sun's Scott McNealy.

Then came the internet: products such as the 8i and 9i databases from Oracle were supposed to solve the problems that came about owing to client-server computing. Riding a wave of popularity prior to the dotcom wave, Oracle reverted to a centralised computing model based around thin clients connecting to back-end servers across the internet.

But such servers faced a major obstacle: to cope with peak demand, users needed to buy far more computing power than they would normally need, leading to vast inefficiencies in terms of under utilisation.

And now for the grid. Companies including IBM, Hewlett-Packard and now Oracle are attempting to tackle this inefficiency by providing computing on demand, where IT workloads can be spread across multiple servers. Users are thus able to run their Oracle database on any server, or group of servers that have spare processing capacity.

It sounds a great idea but there are many barriers. Most operating systems are not robust enough to provide true logical partitioning, which means one application can interfere with another. The second problem is one the industry has yet to tackle: for any model of on-demand computing to work, software needs to be licensed in a standard way, agreed by all major IT manufacturers.

Such control only exists in a proprietary environment, and the only platform that has consistently achieved maximum utilisation is the mainframe, which is a proprietary system.

We have had client-server, the internet and now grid computing: imagine what the IT landscape would be like if we could rub out the last decade of so-called IT progress.

Read more on Business applications