In 2001 IBM’s senior vice president of research, Paul Horn, coined the term Autonomic Computing for computers that are self-configuring, self-optimising, self-healing and self-protecting.
According to David Bartlett, director of autonomic computing at IBM, “Half the problems we are seeing in large customer installations are coming from configuration issues.” Of course, IBM’s definition of large customer is somewhat different from the average Kiwi’s definition of large customer. Even so, the rise of open systems has seen more and more IT departments tempted into an eclectic blend of application servers, database servers and open source applications. The net effect of this is increased complexity of the operational environment. Because of the comparative cheapness of hardware, there is also the tendency to leave current production servers alone and install new solutions onto new servers.
In order to reverse this trend, two of IBM’s initiatives involve more robust and reliable installations and easier problem diagnosis.
IBM and Novell have put forward a new standard for installation packaging to the W3C standards body. It is called Software Installation for Autonomic Computing and it has at the heart of it a new XML scheme for describing solution components, their version dependencies and the actions that should be taken to install them. This standard has been picked up by Installshield and Zero-G and will be incorporated into their standard product offerings.
The target audience is the software developer or installation analyst who will theoretically only need to produce a single installation set for multiple platforms. In practice, the major beneficiaries will be operational staff who will be able to more reliably install new components into complex multi-vendor environments. In terms of a move towards computer autonomy, this can best be described as a modest step.
By contrast the work being done on self-healing is breathtakingly simple in concept and could have a profound long-term effect. The strategy revolves around having a consolidated log in a standard format called the Common Base Event and has been put up to OASIS as an open standard. Basically, you get log adapters that take the log entries from database engines, web servers, routers, custom and package applications and transform them into common base events.
Earlier this year, IBM released the Autonomic Computing Toolkit which has an Eclipse-based iewer Analyser for improved manual issue determination. For larger installations, there is the Tivoli product range including a correlation engine for automatically interpreting and responding to common base events.
In summary, self-configuring systems still seem to be a pipe-dream and getting further away all the time. In contrast, self-healing systems are nearly here and the Autonomic vision is quickly driving us towards a time when we will expect systems to be self-optimising and self-healing.
Reynolds, an independent software consultant in Auckland, travelled to DeveloperWorks courtesy of IBM.