BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20131119T020000Z
DTEND:20131119T040000Z
LOCATION:Booth 3547
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: In most cases, parallel computing I/O comprises a complex dance between the application codes, the compute node based software including MPI and the underlying interconnects.  Until now, the common practice in application programming has been to take the I/O interfaces as a given and to code to them, at some cost to the application in terms of performance. The proposed poster will provide application developers/programmers and the providers of open source middleware architectures and interfaces a basis for discussing whether there are classes of applications that can become higher performance, more efficient and use less power by implementing a new framework of I/O APIs and, possibly, protocols. The prospect is APIs focused on a specific class of applications may achieve significant improvements in I/O efficiency with minimal effort.  The outcome of the discussions at SC13 and thereafter will provide direction to drive the work of the OpenFabrics Alliance going forward.
SUMMARY:Accelerating Improvements in HPC Application I/O Performance and Efficiency
PRIORITY:3
END:VEVENT
END:VCALENDAR