/ 28 October 2011

Crunch time for technology systems

The world of enterprise computing is due for another shake-up in coming years — an inflection point, if you will, that is expected to dramatically change the IT landscape as we know it.

One of the harbingers of this change was the birth of cloud computing, which has shifted computing power, storage and processing to off-site facilities able to leverage combined infrastructure to deliver cost and operational efficiencies.

One of the key challenges that cloud computing addresses is that of provisioning enough storage space to accommodate the ever-increasing volumes of data being generated. The real challenge, however, is to efficiently analyse and archive data in a way that improves business decision-making.

Think like a human
This is the new frontier that companies like IBM are exploring by developing systems that crunch through immense volumes of information to provide an accurate and useful view of the required information. A glimpse into this new landscape was provided earlier this year when IBM unleashed its Watson computer, which runs the company’s DeepQA software that is designed to provide a more human-like understanding of and response to questions. Think HAL 9000 or any other fictional artificial intelligence system popularised in science fiction — although this isn’t science fiction.

In a bid to demonstrate the power of the Watson computer, IBM challenged the producers of the popular American quiz show Jeopardy! to pit their all-time leading contenders against the computer in live, televised competition. The endeavour has close similarities to the 1997 rematch between IBM’s Deep Blue computer and world chess master Garry Kasparov — which Deep Blue won after initially failing in a similar six-match contest the year before.

The victory demonstrated that computers do have the ability to solve problems that had previously been considered the exclusive domain of human intelligence. The Watson computer challenge sought to demonstrate a similar premise – that it is possible to build a system that can operate effectively in human terms rather than strictly in computer terms — with the added fascination that a machine would understand and respond in simple, spoken English.

Understanding is paramount
The key to the Jeopardy! challenge was to produce a computing system able to provide accurate responses within seconds based on clues that could contain subtle meaning, irony, riddles and other natural language complexities. In preparation for the challenge, IBM fed Watson fed four terabytes of content comprising 10 million documents, including all of Wikipedia, the Internet Movie Database, past Jeopardy! clues and the full archive of the New York Times. The entire system was run off a workload-optimised system designed for complex analytics, capable of processing 500 gigabytes per second.

IBM used its DeepQA software, running on the Apache Hadoop framework, which performs natural language processing, reasoning and machine learning tasks. Computing was distributed across 90 IBM Power 750 servers, effectively giving Watson the power of more than 2 280 single-core processor networked computers. As in the 1997 challenge, IBM’s creation came out tops, beating previous champion players Ken Jennings and Brad Rutter in two matches in a record average of three seconds per answer.

“Where Deep Blue was only performing specific operations — analysing chess positions — which amounted to an advanced form of pattern recognition, Watson represents a major leap in the capacity of information technology systems to identify patterns, gain critical insight and enhance decision-making,” says Gary Carroll, director for systems and technology group at IBM South Africa. “The technology holds enormous potential for making the world’s systems smarter.” And it is this opportunity to enhance the speed and accuracy of information mining and delivery from the vast databases and digital storage facilities that promises to redraw the IT landscape.

The Age of Analytics
“The DeepQA technology that powers Watson is, in essence, proof that we are entering a new age in computing — an Age of Analytics. This inflection point will see companies taking a new approach to designing their IT infrastructures to create new opportunities.” says Caroll. This is predicated largely on the correct application of cloud-based computing to offer IT managers flexibility, scalability, processing power and cost efficiencies.

There is a growing list of successful cloud-based implementations that demonstrate exactly how this is possible: from Citigroup that was able to reduce the average turnaround on an application from 45 days to 20 minutes by adopting an internal cloud solution, to the Università di Bari in Italy that implemented an end-to-end cloud-based solution which now provides heavy-duty computing power at minimal cost. Solutions such as these, in conjunction with the breakthroughs heralded by the Jeopardy! challenge, are going to pose new questions of IT managers and their approach to system infrastructure.

Caroll suggests these questions will centre around three key areas. First, the way the enterprise system is accessed, with the emergence of the ubiquitous, mobile device creating massive volumes of information every day. Second, the rise of analytics and the need for real-time insight about and for individual users and, third, the IT architecture that needs to evolve to integrated, flexible infrastructures composed of workload-optimised systems. “By thinking differently about computing, companies can step out of the vicious circle of trying to cater for exploding demand for service on a flat budget and start incorporating big data for better decision making,” he says.

In the mean time, IBM’s Watson computer team is exploring innovative applications for the technology. One area of focus, for instance, is looking at how to use the system to sift through all of the material available to doctors — including newly published research, medical records and case studies — to provide physicians with possible diagnoses and, critically, the evidence of how those potential diagnoses were reached.

This article originally appeared in the Mail & Guardian newspaper as an advertorial