iSGTW Feature - Observing the grid

Observing the grid

Peer through Grid Observatory to explore the world of the grid. Image copyright © 2005-2008 David Opie

What was once lost has now been found-and stored.

Thanks to the Grid Observatory, people in the esoteric field of studying the behavior of large, distributed systems have been given a gift: a trove of data.

Like astronomers who peer through a telescope to explore the solar system, researchers in grids and complex systems are able to examine Grid Observatory's data repository to find new patterns in the global behavior of the grid.

The Grid Observatory, a sub-set of Enabling Grids for E-sciencE's applications group, opened its doors-via a Web portal-this autumn. Through the portal, researchers can access anonymous grid traces, collected and stored through the Grid Observatory application, and information about those traces, for an overall picture of the grid.

Grid traces, explains Cécile Germain of Grid Observatory, are signatures of grid usage. These take many forms: For example, which sites are up, how long a job takes to complete or which scheduling model is being used.

"This information was always collected through EGEE's monitoring tools," says Germain, "now it is stored, archived and can be used in study."

A monitoring tool used by Grid Observatory to visually describe the success of job clusters over several months. Image courtesy of Grid Observatory

Not so esoteric after all

While this field of research is interesting in its own right and is related to studies of other complex systems (such as ant colonies, the economy, nervous systems and the climate) the group also expects that in time it will yield practical benefits as well.

"Before improving a system, you must observe it, says Germain, "then you can strategically place your adjustments. I hope that in time the Grid Observatory will help the grid be more scalable and reliable."

Germain believes that the Grid Observatory will help turn the grid into a complex adaptive system: one that is able to organize itself, change and learn from experience.

"This is goal of autonomic computing (one able to self-manage ) is highly relevant at this time," says Germain, "where production grids are moving to sustainable infrastructures, are experiencing increased usage, and reducing the manpower dedicated to daily operations."

The Grid Observatory is an open project, keen to work with computing researchers.

-Danielle Venton, EGEE