Authors: Andy Bond, John H. Hine
Source: GZipped PostScript (93kb); Adobe PDF
(286kb)
DRUMS is a distributed system used to collect and distribute information about the performance of the individual systems making up the components of a modern network computing facility. DRUMS is designed to achieve a very high level of availability, to have minimum impact on the host systems, to adjust its processing capacity in response to changing demand, and to be easy to manage. Replication is used for both availability and performance. This paper discusses the design and structure of DRUMS and presents results based on our experience with its use.