Feature - OSG 1.0: Stable, Secure, Reliable
Open Science Grid 1.0 is here. Although scientists have been generating results on OSG resources for a couple of years using earlier versions, the VDT-based software has now reached a confident “1.0” level of stability, security and reliability.
A team at the University of Wisconsin, an institution with a long tradition of distributed computing research, leads the OSG software integration, packaging and deployment efforts, and has built all the OSG releases, including 1.0.
After multiple rigorous validation cycles in which stakeholders participated, over a dozen VOs granted the official 'Green Flag' to OSG 1.0.
“OSG 1.0 is an evolutionary, as opposed to revolutionary improvement over the previous software release,” says Alain Roy, software coordinator for OSG. “It provides many updates that improve the stability and reliability of the software and the quality of information available to users and administrators.”
Most important according to Roy are the reliability improvements in collection and dissemination of job statistics to provide a clear picture of real-time operation, and of site-specific parameters and restrictions to help users make decisions about where to send jobs.
“These updates improve the quality and quantity of information OSG can share with the EGEE Grid,” he says. “They are of great interest to the CMS and ATLAS collaborators at the Large Hadron Collider who will rely on this information for daily operation.”
They also enhance the efficiency of the matchmaking software that matches jobs to resources and automatically moves failed jobs to other sites in order to keep work flowing. The matchmaking software and the services it relies on benefit researchers in many disciplines.
Also new in OSG 1.0, sites and users can request and enforce quota-like space leases with lifetime, ownership, and size properties. Opportunistic storage makes its production debut, as OSG works initially with the DZero experiment at Fermilab to use almost a Terabyte per month at each of a half dozen sites.
Among numerous other enhancements, OSG 1.0 works with the OpenSSL version on the resource. This puts security updates for a critical piece of a site’s infrastructure in the capable hands of the operating system vendors, who specialize in tasks like this.
Roy and his team are pleased that since its release in June, over 50% of sites have upgraded to 1.0, and more are following regularly. A site’s local requirements dictate its upgrade schedule; local autonomy is a guiding principle of the OSG grass roots alliance.
Looking ahead, OSG is anxious to move away from updates that require complete re-installation every six months.
“I expect that our future OSG 1.2 release will address the ability to provide incremental updates, and that we will likely stay at that release for a long time, providing these updates as needed,” says Roy.
—Anne Heavey, iSGTW, with Alain Roy, OSG/University of Wisconsin, and Abhishek Singh Rana, UCSD