Net devices evolve toward autonomy

Intelligent networked devices ranging from Internet-enabled embedded microcontroller-based systems to small-footprint Web-enabled smart appliances are experiencing an evolutionary process that can be broken down into three phases: basic connectivity, value-added network functionality and full enterprise membership.

During the first stage, basic connectivity, OEMs have been equipping printers, motor controls, security cameras and the like with essential networking functionality, such as Ethernet 10/100 and basic Internet access. Examining their current competitive landscapes, many electronics OEMs attempt to do this quickly and with as little complexity as possible. Most companies are in or are contemplating this stage.

On the leading edge, a few companies are already well into the Phase 2, value-added stage, where they are coupling the intelligence of a networked microprocessor with networking services. The combination enables data collection and distribution from collaborating devices to enhance the functionality of their products.

The most profound changes will come at Phase 3: enterprise-generated device connectivity. While there are few such devices existing or in design, many companies are considering designs where networked devices become fully participatory and autonomous members of the business operation. Devices become equipped with the necessary networking communications, data and intelligence to make local decisions based on information shared with other devices. For that to happen, MIS systems and devices will pass information seamlessly and automatically-as cooperating partners-without human intervention.

The first hints of such autonomous functionality can be seen in a handful of device applications today, particularly in industrial automation, where the major providers of Web services-particularly IBM, Microsoft and Sun-are expending major efforts.

The enterprise-IT device phase represents the beginning of a new generation of devices and automation services. At present there are few models of Phase 3 applications, but devices using UDDI for lookup and the Extensible Markup Language (XML) and Simple Object Access Protocol (Soap) for communications are coming. A sample application might be a sensor that detects a computer room overheating; the sensor then notifies the air-conditioning system and the MIS department concurrently.

The best net-centric appliances that will fill this space will talk to other devices without human intervention. Devices making decisions and providing online information to back-end systems will change the way business is done, creating a whole new set of services that can be provided. These services will most likely include online diagnostics, predictive maintenance, dynamic function loading, automatic supply ordering, operations rescheduling-and the list goes on.

Based on history, one of the first initiatives that MIS departments will implement is control-over the protocols that the new breed of devices will employ, over their utilization on the corporate network and over which brands can be purchased and deployed.

In order for this to occur, networked electronic products will have to be designed to open standards and for interoperability. Otherwise, they will not be allowed on corporate or commercial networks. Some likely interfaces to consider are DCOM, Corba, Microsoft's .Net and XML with Soap.

As mainframes gave way to minicomputers and minis to PCs, processing shifted to where it was most useful. Tasks were distributed from one central mainframe to several minicomputers, then to hundreds of PCs. In each case, as decision-making moved closer to the user, processing moved closer too, resulting in increased communication between computing systems.

The same shift is happening in the device world. Ultimately, not only will devices be much closer to the decision-making source, they will have the ability to speak to one another. They will also have to communicate with centralized computer systems.

That will require some form of device-to-device communications, most likely according to a form of peer-to-peer communication. Devices are going to have to be able to speak some common language, given that thousands or potentially millions of them will be manufactured by a wide variety of OEMs.

Because of its flexibility and small size, XML is ideally positioned to become the foundation for data exchange. Indeed, XML's rapid adoption by data-intensive application developers proves this point.

With time, schemas will be developed for managing data, databases, control applications, workflow applications and so on. These schemas, and the flexibility of XML, will allow for more-innovative forms of industrial automation and control. For example, cameras will be employed to capture images, extract data from those images, compare it with information in a database and take appropriate action. Inventories will be captured automatically, and processes will be automatically monitored. What's more, these functions will take place in real-time.

The other thing that devices will have to do is cause each other to perform some action. In the past, this has been known as remote log-ins or remote procedure calls. To make distributed processing a reality at the device level, there must be a way for devices to "advertise" their services as well as deliver results. A leading candidate technology for that is the Simple Object Access Protocol, which allows a device to define objects that are, in turn, made available to other devices or for computers to call. Soap runs on top of XML, making it a relatively "small" means for accomplishing the equivalent of remote procedure calls.

With the ability to see how devices can be installed, interchange data and even cooperate in accomplishing an objective, one must still have a central repository for those devices and attributes such that they can be accessed remotely, namely by a directory. The Lightweight Directory Access Protocol has been defined as a way to access such information and is supported by virtually all operating systems. LDAP is relatively small and efficient, making it an appropriate fit for networks.

At the enterprise level, once devices begin communicating in this manner and once systems are established that provide the necessary algorithms for interdevice communication, centralized control is no longer necessary. Automated local control and communications allow even greater efficiencies to be built into these systems. Instead of one massive control program, users have a series of small, distributed programs, allowing programming changes to be isolated and made only where they are needed.

And with decision-making driven down to the device level-the source of the information-the system becomes more configurable, scalable and real-time.

At the hardware and device level, there will be fundamental changes in the kind of processors used and how they are designed. Where 8- and 16-bit devices still have a home in the first and second stages, their days are numbered. Only 32-bit processors have the flexibility to handle data and services running on this type of platform and to deal with two-way data transfers on any kind of busy network.

Additionally, processors in this new segment will be highly integrated system-on-chip devices, with integrated memory, not only for power, but also for performance improvements.

In terms of supporting software, the new platforms will include as essentials in their repertoire support for an XML parser and development tools, Soap and virtual-machine technology, to allow applications to be downloaded and executed seamlessly. Interfaces to MIS middleware like Corba, DCOM and .Net will be need to be considered as their popularity in the embedded world develops. For now, Corba and DCOM appear far too complex for the embedded world, and .Net is too new to judge.