BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:GAON: General-purpose  Application Offload to Near-Network Process
 ors - Mark Silberstein (Technion)
DTSTART:20170914T140000Z
DTEND:20170914T150000Z
UID:TALK73501@talks.cam.ac.uk
CONTACT:Liang Wang
DESCRIPTION:Emerging intelligent bump-in-the-wire network adapters\, iNICs
 \, are increasingly deployed for accelerating custom network functions in 
 large-scale data center and cloud environments such as Microsoft Azure. Ho
 wever\, their potential to speed up network-intensive server applications 
 remains largely unexplored due to the lack of appropriate programming mode
 ls and OS abstractions. \n\nGAON is a framework that facilitates accelerat
 ing network applications using iNICs. GAON introduces an ikernel abstracti
 on that encapsulates the code and state of an iNIC-resident application ta
 sk. The application dynamically invokes the ikernel by attaching it to a n
 etwork socket\, thereby rerouting the socket traffic to the ikernel. GAON 
 can securely execute multiple ikernels of different applications\, isolati
 ng ikernel state and network traffic\, arbitrating access to network from 
 ikernels and CPU applications\, and providing protected access to the iker
 nel state from the owning CPU application. \n\nWe prototype GAON on Mellan
 ox Innova 40 Gbps iNICs and develop several accelerated versions of realis
 tic server workloads. For example\, the transparent key-value store cache\
 nikernel serves 31Mtps (70% of the line rate) at 5 μs 99th percentile lat
 ency for 10 byte key/value pairs\, seamlessly improving the host performan
 ce by up to 55× and 2.3× for 98%\nand 60% cache hit rate respectively\, 
 with almost no modification to the original memcached server running on th
 e same machine.\n\nBio: \nMark Silberstein is an Assistant Professor at th
 e department of Electrical Engineering\, Technion - Israel Institute of Te
 chnology. Mark is interested in accelerator-rich systems and OS abstractio
 ns and services for and on accelerators\, with the ultimate goal to break 
 out of the CPU-centric OS models (HotOS17). His works include GPUfs (ASPLO
 S13)\, GPUnet (OSDI14) and GPUrdma libraries for native file system\, netw
 orking and RDMA support for GPU programs and application-managed virtual m
 emory for GPUs (ISCA16) and SGX enclaves (EuroSYS17).
LOCATION:FW26\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR