• Newsfeed
• Newsletters
• Presentations
• Publications
• Director's letters
• Press

Staged rollout of grid middleware components to EGI

Mário David explains how it works

The EGI staged rollout is a procedure by which the updates of supported grid middleware components, are deployed and tested by Early Adopter sites before made available to all sites integrated in the infrastructure.

The European Grid Infrastructure consists of a few hundred resource centres federated in several dozens of National Grid Initiatives, providing the resources for scientists gathered in Virtual Organisations, also known as VOs. The plethora of different middleware stacks, added to the wide range of scientific disciplines, applications and workflows contribute to the highly heterogeneous environment found in the infrastructure.

The staged rollout process makes it possible to test updates in a production infrastructure, i.e. in environments that are not available on any of the previous testing phases. It allows for potential issues with an update to be discovered at a small scale and workarounds to be added to the release notes if need be. Depending on the severity of the issues or problems, an update may be rejected for wider deployment.

Furthermore, the staged rollout increases the confidence in the quality of the updates that pass limited deployment tests. This guarantees that the vast majority of the sites should experience smooth updates. Resource centres (sites) usually participate in the staged rollout for services that they have a particular interest, and are required to report their findings.

Staging a rollout

In its first year of activities, EGI has established business relationships with two Technology Providers: European Middleware Initiative (EMI), which delivers the gLite, ARC, dCache and UNICORE components, and the Initiative for Globus in Europe (IGE), which provides Globus components.

New versions of middleware components released by Technology Providers are submitted to EGI that processes them through a verification procedure, and after through the staged rollout procedure.

The verification procedure consists in checking a software component against a set of generic criteria, such as available and updated documentation, and specific criteria that depends on the capability under test. It may include deployment in a small test infrastructure, and some level of testing.

Following verification, the staged rollout exposes those newly released versions in the heterogeneous and ‘chaotic’ grid infrastructure. This process was adapted and extended from the previous project Enabling Grids for E-sciencE (EGEE-III), for gLite components.

Over the past one and a half year, the number of Early Adopter has increased to more than 50 sites that have been involved in over 150 staged rollout tests. The rate of rejection of components is low (below 10), but the tests have permitted the discovery of bugs that would otherwise passed unnoticed to the production infrastructure.

The staged rollout testing has contributed to the strengthening, robustness and increased confidence in the middleware that is deployed in the EGI.