MON: On-Demand Overlays for Distributed System Management

This paper presents the management overlay network (MON) system that we are building and running on the PlanetLab testbed. MON is a distributed system designed to facilitate the management of large distributed applications. Toward this goal, MON builds on-demand overlay structures that allow users to execute instant management commands, such as query the current status of the application, or push software updates to all the nodes. The on-demand approach enables MON to be light-weight, requiring minimum amount of resources when no commands are executed. It also frees MON from complex failure repair mechanisms, since no overlay structure is maintained for a prolonged time. MON is currently running on more than 300 nodes on the PlanetLab. Our initial experiments on the PlanetLab show that MON has good performance, both in terms of command response time and achieved bandwidth for software push.