Far from the world of data centers and silicon, an agriculture project at the University of Melbourne is an unlikely place to find the forefront of cloud computing and Web applications.
However, it's a little easier to understand when the university's dean of Engineering, Iven Mareels says profit increases of up to 300 percent on an experimental farm are possible by using irrigation automation in concert with the predictive abilities of IBM's new stream computing software called System S.
With commercialization of the project likely to be four to five years away--available as a cloud computing service rather than as an on-site application--the focus is clearly on the future.
"I envisage several levels of real-time happening: on the one side you have irrigating right now and that is talking time-scales of minutes to hours for the farm; but then you have predictions over the week for a micro-climate; and then you have the longer term of 'what is my market going to be in two months time?'," Mareels told ZDNet Asia's sister site ZDNet Australia last week.
"And bringing that together for many different time scales--and there are actually very few models that transgress time scales of minutes where your [irrigation] controller is acting to months where your decision is acting. To do that all seamlessly and transparently is the trick and it is where we see a big future for stream computing and cloud computing."
In tests on two experimental farms, Mareels says that an orchid farm had a 300 percent increase in profitability in the first two seasons, a dairy farm had an increase of 70 percent and a local commercial diary farmer had similar results, roughly around 70 percent mark.
"The interconnection of the real data that you get in the field, all the way up to the prediction of what the climate looks like, and then feeding that back to the farmer and say 'here is how [to] manage the farm to get maximum profit'-- that's the holy grail," Mareels said.
System S was announced by IBM on May 13, and is built to analyze up to thousands of simultaneous data streams, such as stock prices or weather reports, and process it in real-time.
Glenn Wightwick, an IBM distinguished engineer and director of IBM's Australia Development Laboratory, said the system could scale from an Intel-based blade server up to Blue Gene supercomputers.
The software is model-based and is programmable via the SPADE (Stream Processing Application Declarative Engine) language. But as with any model-based software, it is only as good as the model, which needs expertise to create properly.
"I'm not pretending that it is a magic solution to predicting the global financial crisis, but you could build very sophisticated models with lots of different sources of information that could give you insight into what is going on, but you would need the domain expertise in the creation of those specific models," Wightwick said.