作者: Sudharshan S. Vazhkudai , Douglas Thain , Xiaosong Ma , Vincent W. Freeh
DOI: 10.1109/CLUSTR.2006.311900
关键词:
摘要: Simulations, experiments and observatories are generating a deluge of scientific data. Even more staggering is the ever growing application demand to process assimilate these datasets. Application users perform range data operations, collaborate share in many novel ways. The current storage landscape struggling keep up with trends processing. pay price due over-crowded shared filesystems, or expensive area networks, not enough local storage, high-latency archival wide-area transfers. In order sustain maximize I/O bandwidth relative increasing CPU speeds, applications must take advantage large amounts intermediate commodity However, presents new challenges above beyond traditional distributed file system paradigm: persistent scheduling, storage/CPU coallo-cation, namespace management, lifetime interfaces. this paper, we describe that require suggest several open research problems, illustrate two systems - Freeloader Tactical Storage attack different aspects problems