User Tools

Site Tools


0x13:reports:d3t1t08-conntrack-tales-of-software-datapaths

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
0x13:reports:d3t1t08-conntrack-tales-of-software-datapaths [2019/04/02 12:52] ehalep0x13:reports:d3t1t08-conntrack-tales-of-software-datapaths [2019/09/28 17:04] (current) – external edit 127.0.0.1
Line 1: Line 1:
-Day 3 / Track / Talk 8+Day 3 / Common Track / Talk 8
 Talk – Nuts-n-Bolts: Conntrack: tales of software datapaths Talk – Nuts-n-Bolts: Conntrack: tales of software datapaths
 Speakers: Aaron Conole, Marcelo Leitner Speakers: Aaron Conole, Marcelo Leitner
 Report by: Anjali Singhai Report by: Anjali Singhai
  
-                 i         Lack of offloaded conntrack is visible+Aaron and Marcelo from Redhat begun their talk by providing the motivation for conntrack, being a big DOS vectorHowever conntrack is important. They highlighted that the lack of offloaded conntrack is visible and hinders open vswitch offload, with major cloud vendors abandoning "conntrack" where they can and devising their own connection tracking. Open stack and cloud vendors still need it.
  
-                ii.          Hinders open vswitch offload+Their approach to offload conntrack was that since it must be integrated with current offloading datapath, it requred a software datapath support. Since it was part of netfliter, it was easy to integrate and  integrated with tc.
  
-               iii.          Major cloud vendors are abandoning "conntrack" where they can and devising their own connection tracking.+They showcased the tc actions with connection tracking such as, ct action marking, sending to conntrack, setting zone information. CT action was RCU-fied.
  
-               iv.          Open stack and cloud vendors still need it +As future workthey highlighted that they wanted to better understand performance, provide NAT support and OvS action ct(table=X) creates a copy of the packet and returns the ruleset. How could they do that too? Also offloading hooks still need to be worked out.
- +
-                v.          It must integrate with current offloading datapath. +
- +
-2.      Part of netfliterit’s easy to integrate +
- +
-               vi.          Work done to integrate with tc +
- +
-3.      Ct action marking, sending to conntrack, setting zone information +
- +
-4.      CT action was RCU-fied +
- +
-             vii.          Future work: +
- +
-5.      Better understand performance +
- +
-6.      NAT support +
- +
-7.      OvS action ct(table=X) creates a copy of the packet and returns the ruleset. How can we do that too? +
- +
-8.      Offloading hooks still need to be worked out. +
- +
-9.      Mellanox working with Redhat to get ct work done. +
- +
-10.   Kubernetees need connection tracking…floating IP +
- +
- +
-Site: https://www.netdevconf.org/0x13/session.html?talk-conntrack-tales +
-Slides:  +
-Videos: +
  
 +Mellanox working with Redhat to get ct work done. Kubernetees need connection tracking and floating IP
  
 +Site: https://www.netdevconf.info/0x13/session.html?talk-conntrack-tales
0x13/reports/d3t1t08-conntrack-tales-of-software-datapaths.1554209571.txt.gz · Last modified: 2019/09/28 17:04 (external edit)

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki