On-line application performance monitoring
The primary objective of the CROSSGRID project was to further extend the Grid environment to a variety of applications of great practical importance. Project partners were interested in applications that required a response from the distributed computer system to an action by a user in different time scales (real time or longer time scales). These included interactive simulation and visualisation for surgical procedures, crisis decision support systems and distributed data analysis in high energy physics experiments that are compute- as well as data-intensive. To enable efficient development of this category of applications for the Grid environment, new tools for the verification of parallel source code, performance evaluation and monitoring were needed. More specifically, new services were added for the applications monitoring, while existing services dedicated to monitoring the Grid infrastructure were extended. JIMS was developed by the AGH University of Science and Technology in Kraków to gather and expose information on the state of the devices used to build the Grid environment. The idea was to reuse a standardised Java Management Extensions-based approach to enable monitored devices to expose their operating system parameters to a database system in a unified way. In fact, the acronym JIMS stands for JMX-based infrastructure monitoring system. Moreover, JIMS is a visualisation tool, which allows users to observe the CPU load, RAM and the file system usage statistics diagrams for a chosen worker node, where each application runs. Among the most valuable system features is the ability to dynamically deploy and revoke monitoring agents. The last validated version of JIMS has been installed in a production as well as development testbed and proved to perform efficiently for a growing number of applications.