Distributed applications monitoring at system and network level

Most of the distributed applications are presently based on architectural models that don’t involve real-time knowledge of network status and of their network usage. Moreover the new ”network aware” architectures are still under development and their design is not yet fully defined. We considered, as use case, an application using ODBMS (Objectivity/DB) for distributed analysis of experimental data. The dynamic usage of system and network resources at host and application level have been measured in different client/server configurations, on several LAN and WAN layouts. The aim was to study the application efficiency and behaviour versus the network characteristics and conditions. The most interesting results of LAN and WAN tests are described. The monitoring results identified system bottlenecks and limitations and efficient working conditions in the different scenarios have been defined; some critical behaviours observed when moving away from the optimal working conditions are described. The analysis of the data gathered in the tests have been done off-line. New tools able to visualize on-line the resource usage will give real time information on bottlenecks, which may arise in all the system components, network, host or application itself, and therefore make easier troubleshooting or admission control policy. The current status of tools development is described.