BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Case For In-Network Computing On Demand - Noa Zilberman (Compu
 ter Lab)
DTSTART:20190322T150000Z
DTEND:20190322T160000Z
UID:TALK119002@talks.cam.ac.uk
CONTACT:Marco Caballero
DESCRIPTION:This is a practice talk for EuroSys 2019.\n\nAbstract: \nProgr
 ammable network hardware can run services traditionally deployed on server
 s\, resulting in orders-of-magnitude improvements in performance. Yet\, de
 spite these performance improvements\, network operators remain skeptical 
 of in-network computing. The conventional wisdom is that the operational c
 osts from increased power consumption outweigh any performance benefits. U
 nless in-network computing can justify its costs\, it will be disregarded 
 as yet another academic exercise. In this paper\, we challenge that assump
 tion\, by providing a detailed power analysis of several in-network comput
 ing use cases. Our experiments show that in-network computing can be extre
 mely power-efficient. In fact\, for a single watt\, a software system on c
 ommodity CPU can be improved by a factor of ×100 using FPGA\, and a facto
 r of ×1000 utilizing ASIC implementations. However\, this efficiency depe
 nds on the system load. To address changing workloads\, we propose In-Netw
 ork Computing On Demand\, where services can be dynamically moved between 
 servers and the network. By shifting the placement of services on-demand\,
  data centers can optimize for both performance and power efficiency.  \n\
 nThis is joint work with Yuta Tokusashi (Keio University) \, Huynh Tu Dang
 \, Fernando Pedone\, and Robert Soulé (Università della Svizzera italian
 a).
LOCATION:FW26\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR