We expect problems to arise in complex networked computer systems. Problem determination traditionally requires the presence of highly skilled human problem-solvers, but in autonomic computing (AC) the systems themselves are equipped with some degree of problem diagnosis and resolution capability. Clearly, a natural AC progression is to provide the ability for systems to then self-heal without human input. This article looks at component-oriented development and the role of the humble log file in providing a migration path to automated problem diagnosis and resolution. The standards process is the cornerstone of a successful adoption of autonomic computing.
Service Is King!
Greater reliance on network computing is placing increasingly heavy demands on service providers. Mission-critical systems deployed on networks must have high availability and reliability.
I'm referring here to service provider in the term's broadest sense, including corporate IT departments, outsourcing organizations, and telecom service providers.
Reduced headcount is having the effect of moving more resources off the LAN and into the data center. Outsourcing of IT elements to both external companies and telecom service providers is giving rise to documented service-level agreements. In short, IT users are becoming used to defined (and purchased) levels of service.
Telecom service providers are working hard to cope with reduced revenues from traditional services, such as voice and leased lines. We now see service providers offering virtual private networking (VPNs), web and content hosting, security management, and so onall in an effort to raise much-needed revenue.
The net effect of this change is networks that are denser and more complex. So great is the complexity that it's doubtful human operators will be able to rise to the management challenge unaided. To illustrate the issues, let's imagine that a service provider offers 10 services to a population of 100 corporate customers. This setup is just about manageable given existing technology. However, if the number of services is increased to 1,000 and the number of customers grows to 10,000, there is simply too much complexity in the network. In the area of common problem resolution, human input needs to be reduced.