Analysis of start-up transition dynamics of TCP NewReno

C. Joo, S. Bahk

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)

Abstract

Internet traffic rapidly grows in volume and mainly comes from TCP-related application services. Therefore, many TCP variants have been researched actively for a long time and among these, TCP NewReno has been proposed to improve TCP performance by preventing unnecessary timeout, which results in multiple packet losses. NewReno successfully recovers multiple packet losses without any changes at receivers and removes unnecessary timeout. However, since NewReno can recover only a packet during one RTT, it may take a long time to recover all packet losses and it still suffers performance degradation, especially when it experiences multiple packet losses during its start-up period. As the network bandwidth grows and applications that use short-lived TCP connections increase, TCP in the start-up period plays a major role in data transfer, and significantly contributes to the overall performance. In this paper, we focus on the importance of start-up transition dynamics of TCP NewReno's with and without partial window deflation (PWD). They have different window reduction mechanisms for recovery when a loss is detected. We analyze them in detail and verify our analysis through extensive simulations. Among the recovery procedures that are linked with timeout, Fast Retransmit, and Fast Recovery, our work is primarily concerned with the dynamics of Fast Recovery.

Original languageEnglish
Pages (from-to)237-250
Number of pages14
JournalComputer Networks
Volume36
Issue number2-3
DOIs
Publication statusPublished - 2001 Jul
Externally publishedYes

Keywords

  • Fast recovery
  • Partial and duplicate ACKs
  • Partial window deflation (PWD)
  • Start-up dynamics
  • TCP NewReno

ASJC Scopus subject areas

  • Computer Networks and Communications

Fingerprint

Dive into the research topics of 'Analysis of start-up transition dynamics of TCP NewReno'. Together they form a unique fingerprint.

Cite this