The dramatic growth of computer networks creates both an opportunity and a daunting distributed computing problem for users seeking to build applications that can configure themselves and adapt as disruptions occur. The problem is that data often resides on large numbers of devices and evolves rapidly. Systems that collect data at a single location scale poorly and suffer from single-point failures. Here, we discuss the use of a new system, Astrolabe, to automate self-configuration, monitoring, and to control adaptation. Astrolabe operates by creating a virtual system-wide hierarchical database, which evolves as the underlying information changes. Astrolabe is secure, robust under a wide range of failure and attack scenarios, and imposes low loads even under stress.
Kenneth P. Birman, Robbert van Renesse, James Kauf