Max-Planck-Institut für Informatik
max planck institut
mpii logo Minerva of the Max Planck Society

MPI-INF or MPI-SWS or Local Campus Event Calendar

<< Previous Entry Next Entry >> New Event Entry Edit this Entry Login to DB (to update, delete)
What and Who
Title:High Performance Data Center TCP Packet Processing
Speaker:Antoine Kaufmann
coming from:University of Washington
Speakers Bio:Antoine Kaufmann is a Ph.D. candidate in Computer Science and Engineering at the University of Washington, where he is a member of the Computer Systems Lab. Previously, he completed a Master's Degree from the Swiss Federal Institute of Technology (ETH) in Zurich. Antoine's research area is operating systems and networks, and much of his recent work has focused on improving application I/O performance in the data center.
Event Type:SWS Colloquium
Visibility:D1, D2, D3, D4, D5, SWS, RG1, MMCI
We use this to send out email in the morning.
Level:AG Audience
Date, Time and Location
Date:Monday, 19 February 2018
Duration:60 Minutes
TCP is widely used for client-server communication in modern data centers. But TCP packet handling is notoriously CPU intensive, accounting for an increasing fraction of data center processing time. Techniques such as TCP segment offload, kernel bypass, and RDMA are of limited benefit for the typical small, frequent RPCs. These techniques can also compromise protocol agility, resource isolation, overall system reliability, and complicate multi-tenancy.

I propose a unique refactoring of TCP functionality that splits processing between a streamlined fast path for common operations, and an out-of-band slow path. Protocol processing executes in the kernel on dedicated cores that enforce correctness and resource isolation. Applications asynchronously communicate with the kernel through event queues, improving parallelism and cache utilization. I show that my approach can increase RPC throughput by up to 4.1x compared to Linux. The fast-path can be offloaded to a programmable NIC to further improve performance and minimize CPU time for network processing. With hardware offload, data packets are delivered directly from application to application, while the NIC and kernel cooperate to enforce correctness and resource isolation. I show that hardware offload can increase per-core packet throughput by 10.7x compared to the Linux kernel TCP implementation.
Name(s):Susanne Girard
EMail:--email address not disclosed on the web
Video Broadcast
Video Broadcast:YesTo Location:Saarbr├╝cken
To Building:E1 5To Room:029
Meeting ID:
Tags, Category, Keywords and additional notes
Attachments, File(s):
Susanne Girard/MPI-SWS, 02/13/2018 11:36 AM
Last modified:
Uwe Brahm/MPII/DE, 02/19/2018 07:01 AM
  • Susanne Girard, 02/13/2018 11:46 AM -- Created document.