Australian IT 25 July 2006

Open path no threat to majors

Eric Wilson
July 25, 2006

OPEN source software is a small but emerging presence, according to a Gartner Business Intelligence hype cycle report issued last month.

Some leading Australian practitioners disagree with the conventional wisdom that free BI still faces a long road to maturity.

Institute of Analytics Professionals Australia ACT chapter head Graham Williams admits that open source's quality is often poor but it doesn't stay that way.

"As people use it, the quality quickly surpasses commercial products," Williams says. "Some of the applications evolve quickly into a high-quality use of the technique.

"People need to take into account that free BI attracts the latest statistical techniques way before the commercial vendors," he says.

"This is because academics find it easier to integrate new ideas into an open-source project than to convince vendors to adopt them.

"So if a new technique proves valuable, its bugs are quickly ironed out by the corporate developers adopting it."

Williams has used an open source data-mining tool called R in more than 50 projects for a federal government department.

Association chair Inna Kolyshkina says leading academics are behind R, but she does not knows of any official training courses.

"This means R requires strong programming and analytical skills," she says.

That's why Bill Hostmann, a Gartner research vice-president, describes open source as on the edge of BI users' radar screens.

"People are using it to embed components such as statistical graphics into their applications," Hostmann says.

Geoff Peach, chief technology officer at Data Warehousing Services Australia, gives away such free software carrots to entice users to employ his company's services, but he hasn't found any free BI that's of sufficiently high standard.

"I believe it's coming," Peach says.

People are going to start looking at their data through free tools.

One of these is BIRT, a plug-in for the open source Eclipse software development system. The resulting statistical models can then be run on a proprietary Actuape platform.

A data-mining project can be started for free but it will need commercial software to scale up.

"There's a lot of developer interest in the code libraries," Hostmann says, but it's too early to tell if this hybrid approach will work.

"Sometimes you can't print out results in a certain way because the vendor has locked you in," Williams says. "With open source you can always get at your data because it's never in a proprietary format.

"We can get it exactly the way we want it, which is very powerful.

"With commercial vendors, improvements often don't appear until the next update," Williams says.

Yet R suffered from not having a graphical user interface, so Williams rolled his own, which is now used by hundreds of others.

Peach says companies such as Cognos and SAS will eventually be affected by free software.

"You won't have to pay $500,000 for a BI tool," Peach says.

Hostmann predicts prices will plunge long before open source bites the proprietary incumbents, as Microsoft enters the market.

By way of analogy, open source MySQL hasn't taken out Oracle's core database business, Hostmann says.

Likewise, free software will probably expand the business intelligence community, not replace its major vendors.

The Australian

Ads Follow - These are Not Endorsed by Togaware
Shop at Amazon