BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20101118T193000Z
DTEND:20101118T210000Z
LOCATION:384-385
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: The need for parallel IO is becoming ever more pressing - With several libraries available, some more established than others, and some seemingly filling niche roles for particular scientific domains, the developer has a difficult time choosing the right library, and getting the best performance from it. =0AIf parallel file systems can be scaled up to support more and more IO nodes and metadata servers, then is the issue of raw bandwidth a 'solved' problem? If so, how can library developers ensure that scientific applications get the maximum performance without requiring extensive understanding of the underlying methods and issues.=0A=0AIn this panel, we bring together developers of some of the leading parallel IO libraries used in the scientific computing community and ask how they can be best used by developers (such as climate modelers) producing prodigious amounts of data and how will the libraries change to meet future needs.
SUMMARY:Parallel I/O: Libraries and Applications, Making the Most of Resources
PRIORITY:3
END:VEVENT
END:VCALENDAR