Outside the workplace, it isn't computers that count, but what are known as pervasive devices. Telephones, mobiles, digital TVs, PDAs (Personal Digital Assistant), point-of-sales (POS) terminals in shops and cash dispensers are all examples of pervasive devices in that they are special-purpose items widely distributed throughout the population which incorporate some form of IT.
Pervasive devices matter because they offer the smart business a way of amassing detailed information on exactly how customers behave. The idea that successful Web sites have to collect information on visitor behaviour and use it for business advantage is already widely accepted. Businesses now need to realise they can achieve similar benefits by collecting information on customer behaviour from pervasive devices for use in real-time. An elementary attempt at capturing and responding to customer behaviour is already being undertaken by some supermarkets, which offer customers vouchers at the checkout promoting items similar to those they have already purchased.
The other part of the business equation is that pervasive devices, mobiles in particular, represent a costly systems management headache. Any widely distributed device is a logistical nightmare to repair or upgrade, so the ability to diagnose faults and apply solutions remotely via software agents is worth a great deal.
Tivoli Systems claims to have just such agents together with the ability to integrate agent data into mainstream business systems. Tivoli was formed in 1989 to develop and sell IT systems management facilities. It has done this by creating an agent framework, called Tivoli Management Framework, plus agents to collect data, applications defining the data to be collected and 'managers' to run the framework. In 1996, the company was acquired by IBM and now operates as an autonomous division. Tivoli claims its systems management technology is used by 95% of the global top 1,000 companies.
By themselves agents are of little use. They need a framework to allow them to intercommunicate and aggregate, analyse and manipulate the data they collect. But frameworks and agents are expensive items, take time to implement and are unlikely to be cost-effective if all they do is collect additional customer information from pervasive devices. However, if a company is already using a framework for systems management, it doesn't need another for collecting customer information. All it requires is a new agent and a feed into customer information databases.
Tivoli recently extended its IT management framework and claims to have agents now for the Palm PDA (Windows CE PDA support will come later this year), POS terminals, cash dispensers, cable modems and digital set-top boxes. These are in various states of commercial readiness. The Palm PDA agent, for instance, can communicate with a wider infrastructure only if physically connected to a PC.
The real need is for wireless communication, enhanced or additional agents, and agents for more platforms. But Tivoli will have to partner well because that kind of development will cost a fortune. Only then will the company be able to realise the potential of its principal contribution - the integration of information flows with mainstream IT systems.
The fast-growing market for digital TV set-top boxes is a good example of the growth of pervasive devices.
Agents here would support remote changes to programme entitlements and the collection of programme preference information.
Then there's mobile phones and PDAs. The number of mobile phones in use worldwide is expected to exceed a billion in three years. Growth in PDAs may be less evident but is similarly spectacular - in 1999 Gartner put the market base at eight million and estimated it would increase to 38 million by 2003.
Palm dominates the PDA market with a 75-80% share. Windows CE/PocketPC devices produced by big-name suppliers such as HP, Compaq and Casio have a small but significant share. The mobile phone market is split between global players such as Motorola, Philips and Nokia. Tivoli's agents for Palm PDAs put the company in a good position if PDAs replace laptops as business tools, and if managers want control of the devices and of the personnel they help.
Keeping the auditors happy may be justification enough for corporates to roll out pervasive device management. Auditors are increasingly conscious of the growth of distributed IT assets and unlikely to accept a statement that an organisation has 10,000 or so desktops/laptops 'out there somewhere' as evidence of adequate control. They'll want an inventory of who has them, where they are and what they are worth. As the population of PDAs grows, so their cost will figure more prominently in company accounts and it's easy to anticipate auditors adopting a similar attitude to PDAs as they currently do to PCs. If a company can't demonstrate it has control of its mobile assets, it's going to find rude notes left on the accounts.
An insight into the benefits of control of distributed (and mobile) assets was recently provided by a Mexican truck company. The firm had invested in global positioning systems for its trucks so that it could trace their progress on deliveries. Its investment brought down average delivery times from 3.5 hours to 20 minutes.
Closer to home, the Safeway supermarket chain purchased Tivoli agents (and supporting software) for POS terminals at its 470 stores in the spring of this year. The aim was to reduce stock levels by turning overnight batch re-ordering from store controller aggregations into a real-time process based on reports from the terminals themselves.
Tivoli doesn't have this market to itself. Suppliers will want to create their own agents if they add value to their products. ICL has a stake in the retail POS market and existing agent technology.
But Tivoli scores with its existing agent framework that drives through mid-tier servers to back-office information systems. And the ability to integrate data collected from pervasive devices into mainstream business systems is a huge asset. Creating an agent framework from scratch has been a massive investment for Tivoli. Other suppliers are unlikely to want to repeat that investment but highly likely to produce their own agents.
Businesses know they must integrate pervasive devices into mainstream systems. Look at the widespread but flimsily founded interest in Windows CE fuelled by promises of seamless integration with Windows on the desktop although it is, in fact, a quite different animal to its bigger desktop cousin.
Much depends on Tivoli's ability to persuade device suppliers to integrate their agents with its infrastructure. If successful, this would also relieve Tivoli of the need to invest itself in developing agents for the disparate pervasive devices already in existence and whose diversity is likely to increase in the future.
Timescales are key to the success of Tivoli's vision. The pervasive device market is moving extremely fast and the technical developments Tivoli still has to undertake may not take long but must be constrained by its budget. One way or another, big technical advances are very likely within three years. How well Tivoli fares will depend less on its technical abilities than on how well it overcomes the investment barrier.
Agent technology is well understood and proven. There are no obvious technical barriers to Tivoli's plans and no reason to assume the company won't be able to implement the necessary technical enhancements.
Tivoli's IT infrastructure management framework and existing user base are the prime assets the company seeks to exploit. The test for Tivoli will be how well it can grow with the rapidly expanding market. It can't expect to do so successfully on its own resources, even with backing of IBM, so the key lies in the success or failure of its partnership programme.
In the Pipeline
Tivoli's main development centres are in Texas, California and Rome. The company specialises in remote, centralised control of IT and IT-enabled devices, so its research focus is limited.
The general directions of research can be described in the form of a matrix, with identification of 'application bundles' for specific industries as the vertical axis and types of device as the horizontal axis.
For the purposes of this Lab Report, agents are best thought of as small software entities that reside on a particular device at any one time and collect information on what is happening on that device.
Agents have many similarities with viruses and are sometimes referred to as 'benign viruses'. They also share characteristics with objects: they communicate (brokerage) within a framework, and make requests and respond to requests from other agents (invoke methods).
Agent technology first came to the fore about 12 years ago and, like most brave new worlds in IT, failed to live up to the early hype. Part of the problem is that many industry-wide standards for distributed computing (beyond Java applets) would need to be created and implemented before the dream of globally interoperable agents could be realised.
Nonetheless, in specific application domains and where common technology is widespread, agents can fulfil very useful roles more effectively than the alternatives.