Indiana University Dedicates New Resources to Support Lustre over WAN

University Information Technology Services (UITS) at Indiana University announced today that it has dedicated over 350 terabytes (TB) of new storage platforms to support collaborative research projects mounting the Lustre file system across the TeraGrid network and other national high speed networks. IU's Data Capacitor Project Lead Stephen Simms made the announcement during a panel discussion at the Sun Microsystems-sponsored 2008 Lustre User Group being held this week in Sonoma, California. The ability to use Lustre over a wide area network (WAN) is a significant advancement in the ongoing struggle to meet user demand for easier and faster access to stored research data generated by high performance computers. "A large impediment to maintaining a wide area file system that spans multiple organizations is the need for a unified user identification scheme across all clients," said Simms, who also serves as IU's TeraGrid site lead. Through small modifications to the Lustre code base, Indiana University — with help from Data Direct Networks — has recently developed a lightweight scheme for mapping user IDs (UIDs) across multiple distributed clients, ensuring that file ownership permissions will remain consistent across all clients. This development is a critical step toward widespread use of Lustre as a wide area file system, and will provide greater security and ease of use. Early Lustre WAN projects required strict synchronization of UIDs between client systems; IU's UID mapping scheme eliminates the need for this strict synchronization. Implementation of this new code will grant appropriate access to only those client UIDs that have been registered with the Lustre administrator. Indiana University has made several notable achievements related to the use of Lustre over a WAN in the past year. In June, IU announced that its Data Capacitor — a 535 TB Lustre file system designed to store and manipulate large data sets — demonstrated in its opening weeks of production a single client transfer rate of 977 MB per second across the TeraGrid network. In November, a team led by Indiana University was awarded first place in the annual Bandwidth Challenge at the SC07 Conference in Reno, Nevada; using the Data Capacitor, the IU team achieved a peak transfer rate of 18.21 Gigabits/second out of a possible 20 Gigabits/second. This performance was nearly twice the peak rate of the nearest competitor. "Indiana University has had amazing success using Lustre in the last several months," said Simms. "We've shown how the Data Capacitor allows researchers to access, manage, and manipulate massive data sets quickly and easily from remote locations. We are very excited to extend that work and offer a new resource dedicated to wide area collaborations." The Data Capacitor project is supported in part by the National Science Foundation under NSF Award Number CNS0521433 (Craig Stewart, PI; Stephen Simms, Co-PI and project manager; Caty Pilachowski, Randall Bramley and Beth Plale, Co-PIs). IU's involvement in the TeraGrid is supported in part by NSF grants ACI-0338618l, OCI-0451237, OCI-0535258, and OCI-0504075.