MonALISA monitoring.
Image Courtesy Iosif Legrand |
Operating a successful grid, network or computing facility requires vast amounts of monitoring information. Projects and organizations worldwide that need to track resource usage, network traffic, job distribution and many other quantities rely on Caltech's MonALISA system to collect the information and present it in a way that allows them to make effective decisions. The system also automatically troubleshoots and optimizes very large grid and network systems.
"While developing computing models for experiments at the Large Hadron Collider, we realized that vital monitoring information was missing," explained Caltech's Iosif Legrand, MonALISA's chief architect. "We didn't want to reinvent the wheel, so we built a framework that collects and synthesizes data from many existing monitoring tools and built new tools only where needed."
MonALISA, which stands for Monitoring Agents Using a Large Integrated Services Architecture, has been developed by Caltech and its partners over the last five years. The framework monitors everything from single computers to large facilities and global networks. At CERN in Geneva, Switzerland, MonALISA services monitor 2,500 computers and transcontinental and intercontinental network links. The Open Science Grid uses the framework for resource and network monitoring, and the GLORIAD network monitors traffic and all major routers using MonALISA.
Legrand and a team of graduate students from the Polytechnic University of Bucharest in Romania develop and maintain the framework, which relies on four layers of services: lookup discovery services; MonALISA services and mobile agents; proxies; and repositories. The MonALISA services register themselves with the discovery services and deploy mobile agents which locally extract or create all types of monitoring information. Proxies gather the information from the MonALISA services and send it to clients who have requested it. Repositories provide a central location for users to find any piece of information from an entire system. New client-based agents called LISA profile the capabilities and state of users' computers and the quality and performance of their network connections.
"We're now also starting to integrate accounting information into the framework," said Legrand. "In the future we will make it easier to analyze the increasing amount of information being collected, develop higher-level services and provide an optimized workflow into large grid systems."
Learn more at the MonALISA Web site.
—Katie Yurkewicz
e-mail this article
|
