Name: A new slow start algorithm for transport layer protocol (TCP and QUIC)

Description

TCP slow start is designed to ramp up to the network congestion point quickly,
doubling the congestion window each round-trip time until link capacity is
reached, whereupon TCP exits the slow start phase. Unfortunately, the default
Linux TCP slow start implementation -- TCP Cubic with Hystart
-- can cause premature exit from slow start, especially over wireless links,
degrading link utilization. However, without HyStart, TCP exits too late,
causing unnecessary packet loss. To improve TCP performance during slow start,
This document proposes the Slow start Exit At Right CHokepoint (SEARCH)
algorithm where the congestion point is determined
based on bytes delivered compared to the expected bytes delivered, smoothed
to account for link latency variation and normalized to accommodate link
capacities. We request a time slot in IETF 120 for discussion on this algorithm.

Required Details

• Status: (not) WG Forming
• Responsible AD:
• BOF proponents: Jae Won Chung <jaewon.chung@viasat.com>,Maryam Ataei Kachooei <mataeikachooei@wpi.edu>, Feng Li <feng.li@viasat.com>, Mark Claypool <claypool@wpi.edu>
• Number of people expected to attend: 50
• Length of session (1 or 2 hours): 2 hours
• Conflicts (whole Areas and/or WGs)
• Chair Conflicts: n/a
• Technology Overlap: n/a
• Key Participant Conflict: n/a

Information for IAB/IESG

To allow evaluation of your proposal, please include the following items:

• Any protocols or practices that already exist in this space:
  Hystart and Hystart++, however, neither works well over large BDP networks.

• Which (if any) modifications to existing protocols or practices are required:
  TCP CUBIC (etc) and QUIC

• Which (if any) entirely new protocols or practices are required:
  N/A

• Open source projects (if any) implementing this work:
  starting a pull request (PR) on H2O/quicly

Agenda

• Items, Internet-Drafts, speakers, timing
• Or a URL