BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111116T223000Z
DTEND:20111116T230000Z
LOCATION:WSCC 611/612
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Kernel-level software-based Symmetric Multi-Processing (SMP) supercomputing clusters of up to 3 homogeneous server nodes were introduced in 2011.  With the new 16-core AMD Interlagos processors, SMP supercomputers with 192 cores and 1.54 TB of shared memory are now available for well under €100K.  Next year, kernel-level software-based SMP supercomputing clusters of 32 nodes will be introduced – that is SMP supercomputers with 2048 cores and 16 TB of shared memory for under €1 million.=0A=0ADuring this session, we  will explore key issues for kernel-level software-based SMP supercomputing.  How global memory speed and memory coherency is maintained across multiple homogeneous server nodes.  How the advent of high-speed interconnects between homogeneous server nodes in a cluster enables SMP computing.  And, why kernel-level software-based SMP supercomputing is the next logical evolution to proprietary hardware and hypervisor software approaches to SMP supercomputing.
SUMMARY:High Performance SMP on Clusters with Kernel-Level Software
PRIORITY:3
END:VEVENT
END:VCALENDAR