Grenville Armitage and Gorry Fairhurst TCPM WG - IETF 94, Yokohama 5 - - PowerPoint PPT Presentation

Update on TCP Alterna0ve Backoff with ECN (ABE) dra=-khademi-alterna0vebackoff-ecn

Naeem Khademi, Michael Welzl, Grenville Armitage and Gorry Fairhurst

TCPM WG - IETF 94, Yokohama – 5 November 2015

• An experimental update to RFC3168: updates the TCP sender-side

reacMon to a congesMon noMficaMon received via ECN

• RFC3168 Sec0on 6.1.2:

“If the sender receives an ECN-Echo (ECE) ACK packet…the indicaMon of congesMon should be treated just as a conges0on loss in non-ECN-Capable

• TCP. That is, the TCP source halves the congesMon window cwnd and reduces

the slow start threshold ssthresh.”

• Update to RFC3168 Sec0on 6.1.2:

“If the sender receives an ECN-Echo (ECE) ACK packet…the indicaMon of congesMon SHOULD induce a less conserva0ve reac0on than loss: the TCP source mulMplies the congesMon window cwnd with 0.8 and reduces the slow start threshold ssthresh.”

I-D’s Scope

2

• Discussions on the TCPM ML (up to IETF Prague):

– Addressed Mark Allman’s comments with regards to betaecn as a percentage – Review from Anil Agarwal is responded

• Discussions on the ICCRG ML (a=er IETF Prague):

– Responded to comments from Michael Scharf and Bob Briscoe

• SubmiXed -01:

– Some editorial work, more elaboraMve text on the raMonale behind ABE in IntroducMon (SecMon 1) and Discussion (SecMon 2) – Discussion (SecMon 2) explains:

• where ABE can be useful (lightly mulMplexed, high-BDP access links)
• ABE is prac,cal rather than ideal (staMc beta rather than an adapMve one)
• The choice of mulMplier (no change to betaloss)

– should -> SHOULD in the update of RFC3168 (SecMon 3) – The proposed update is “experimental” rather than standard-track

Status of the I-D

3

Q&A

4