Feature: GLUE for the Grid
A defining feature of a grid is that its components can change from minute to minute: storage, processors and sites join and leave, jobs start and finish and files are written and deleted. Resources can also vary widely, from disk to tape storage, different operating systems and processor types.
But for the grid to work, the status and makeup of each of these contributions needs to be known, so that resources can be allocated to users, sites monitored and accounting data collected. For many of the world's scientific grids, this problem is solved by the Grid Laboratory Uniform Environment-or GLUE Schema. Many projects, such as Enabling Grids for E-sciencE, the Worldwide LHC Computing Grid and the Open Science Grid, use GLUE.
The GLUE Schema is a common way of publishing information, developed by consortium of grid projects, including EGEE, OSG, the Australian Partnership for Advanced Computing and the Nordic DataGrid Facility. It describes attributes of grid sites and services, computing elements and storage elements. Implementations of the GLUE Schema exist for a range of systems such as the LDAP directory access system, R-GMA monitoring system and XML language.
"The GLUE Schema is a good example of cooperation between grid projects," says Stephen Burke from the UK's GridPP who has been working on the definition and evolution of the schema. "It allows them to interoperate by representing the state of the grid services and resources in a uniform way."
Initially promoted by the transatlantic research project DataTAG and International Virtual Data Grid Laboratory, GLUE was later adopted by many major grid projects.
"It is deployed in OSG to support grid monitoring, interoperability with the LCG and a resource selection service," says Shaowen Wang from the University of Iowa. "Our experience in OSG has shown that grid information modeling work like the GLUE Schema is a fundamental element to enable scalable federation of large-scale production-quality grids."
Last October, the GLUE schema collaboration held a meeting at Imperial College, London, to plan for new versions.
"The latest schema revision," Burke says, referring to version 1.3, "will allow the greater flexibility of the new SRM 2.2 storage protocol to be represented, and also resolves some outstanding problems, while remaining backward-compatible with the previous version."
The collaboration also began discussing a major revision of the GLUE schema, Version 2.0. To develop the plans a new working group will be proposed during the 19th Open Grid Forum in Chapel Hill, North Carolina. The goal is to have a new schema defined by the end of 2007.
"The proposed formation of a GLUE working group in OGF should enable the participation of more projects, and lead to a generally accepted standard schema," adds Burke.
Implementations of the schema, version 1.3, are expected to be deployed early this year, and details can be found on the GLUE Web site.
- Sarah Pearce, GridPP