US research could lose lustre

Lustre Technical White Paper

Scaling Spark on Lustre

We report our experiences in porting and tuning the Apache Spark data analytics framework on the Cray XC30 (Edison) and XC40 (Cori) systems, installed at NERSC. We find that design decisions made in the development of Spark are based on the assumption that Spark is constrained primarily by network latency, and that disk I/O is comparatively cheap. These assumptions are not valid on Edison or Co...

Distributed Lustre activity tracking

Numerous administration tools and techniques require near real time vision of the activity occuring on a distributed filesystem. The changelog facility provided by Lustre to address this need suffers limitations in terms of scalability and flexibility. We have been working on reducing those limitations by enhancing Lustre itself and developing external tools such as Lustre ChangeLog Aggregate a...

Temporal Refinement for Lustre

This paper proposes a refinement calculus for Lustre. First a very general calculus is provided, which ensures correctness and reactivity for a large class of systems. Then, this calculus is adapted to provide oversampling and temporal refinement. We obtain thus an effective calculus for Lustre, which allows us to refine both computations and time. We illustrate its use on a small example and c...

CLINICAL-PATIENT STUDIES I could lose everything: understanding the cost of a brain tumor

Although studies have quantified the costs of cancer treatment, few have evaluated the widespread impact of cancer costs on the family unit. Specifically, little is known regarding how cancer affects patients and their families financially, and how they cope with these costs. The purpose of this descriptive study was to explore the financial impact of cancer care in neuro-oncology. Content anal...