Self-Managed Computer Systems: Foundations And Examples

The traditional approach to managing complex computer systems is to use a cadre of skilled IT professionals who use monitoring tools in order to detect when problems arise. They are then able to use their skills and experience to determine what actions should be taken to solve the problems. This approach is no longer viable for highly complex, networked computer information systems that have numerous configuration knobs, and operate in environments that vary with time at a very high rate. In this case, one cannot expect that design-time configurations will make the system operate optimally at run-time. For that reason, complex systems need to manage themselves using controllers that make the systems self-configuring, self-optimizing, self-healing, and self-protecting. This paper provides a formalism to describe self-managed systems and discusses concrete examples that illustrate how these properties are enforced by controllers in a variety of domains including cloud computing, fog/cloud computing, Internet datacenters, distributed software systems, and secure database systems.