BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:2.0
BEGIN:VEVENT
DTSTART:20151117T231500Z
DTEND:20151118T010000Z
LOCATION:Level 4 - Lobby
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: rCUDA is a state-of-the-art remote CUDA execution middleware that enables CUDA=0Aapplications running on one node to transparently use GPUs on other nodes. With=0Athis capability, applications can use nodes that do not have enough unoccupied=0AGPUs by using rCUDA to borrow idle GPUs from some other nodes. However, those=0Aapplications may suffer from rCUDA's overhead; especially for applications that=0Afrequently call CUDA kernels or have to transfer a lot of data, rCUDA's=0Aoverhead can be detrimentally large. We propose mrCUDA, a middleware for=0Atransparently live-migrating CUDA execution from remote to local GPUs, and show=0Athat mrCUDA's overhead is negligibly small compared with rCUDA's overhead.=0AHence, mrCUDA enables applications to run on nodes that does not have enough=0Aunoccupied GPUs (by using rCUDA) and later migrate the work to local GPUs=0A(thus, get rid of rCUDA's overhead) when available.
SUMMARY:mrCUDA: Low-Overhead Middleware for Transparently Migrating CUDA Execution from Remote to Local GPUs
PRIORITY:3
END:VEVENT
END:VCALENDAR