Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
TCP-FIT: An Improved TCP Congestion Control
Search
Sponsored
·
Ship Features Fearlessly
Turn features on and off without deploys. Used by thousands of Ruby developers.
→
Kevin Tong
May 13, 2013
Technology
0
510
TCP-FIT: An Improved TCP Congestion Control
Kevin Tong
May 13, 2013
Tweet
Share
More Decks by Kevin Tong
See All by Kevin Tong
Improving Fairness, Efficiency, and Stability in HTTP-based Adaptive Video Streaming with FESTIVE
logicmd
1
4.6k
Transport Methods in 3DTV—A Survey
logicmd
0
140
The Performance of MapReduce: An In-depth Study
logicmd
0
690
Simple Regenerating Codes:Network Coding for Cloud Storage
logicmd
0
520
ANALYSIS OF ADAPTIVE STREAMING FOR HYBRID CDN/P2P LIVE VIDEO SYSTEMS
logicmd
1
540
Other Decks in Technology
See All in Technology
「Blue Team Labs Online」入門 - みんなで挑むログ解析バトル
v_avenger
0
180
Claude Code Skills 勉強会 (DevelersIO向けに調整済み) / claude code skills for devio
masahirokawahara
1
21k
The_Evolution_of_Bits_AI_SRE.pdf
nulabinc
PRO
0
220
脳内メモリ、思ったより揮発性だった
koutorino
0
360
JAWSDAYS2026 [C02] 楽しく学ぼう!AWSとは?AWSの歴史 入門
hiragahh
0
160
GCASアップデート(202601-202603)
techniczna
0
140
マルチアカウント環境でSecurity Hubの運用!導入の苦労とポイント / JAWS DAYS 2026
genda
0
710
アーキテクチャモダナイゼーションを実現する組織
satohjohn
2
930
JAWSDAYS2026_A-6_現場SEが語る 回せるセキュリティ運用~設計で可視化、AIで加速する「楽に回る」運用設計のコツ~
shoki_hata
0
3k
20260311 技術SWG活動報告(デジタルアイデンティティ人材育成推進WG Ph2 活動報告会)
oidfj
0
350
Google系サービスで文字起こしから勝手にカレンダーを埋めるエージェントを作った話
risatube
0
190
VLAモデル構築のための AIロボット向け模倣学習キット
kmatsuiugo
0
160
Featured
See All Featured
Context Engineering - Making Every Token Count
addyosmani
9
750
How to optimise 3,500 product descriptions for ecommerce in one day using ChatGPT
katarinadahlin
PRO
1
3.5k
Designing Dashboards & Data Visualisations in Web Apps
destraynor
231
54k
Visualization
eitanlees
150
17k
Why You Should Never Use an ORM
jnunemaker
PRO
61
9.8k
Measuring & Analyzing Core Web Vitals
bluesmoon
9
780
Documentation Writing (for coders)
carmenintech
77
5.3k
Navigating Team Friction
lara
192
16k
Kristin Tynski - Automating Marketing Tasks With AI
techseoconnect
PRO
0
190
VelocityConf: Rendering Performance Case Studies
addyosmani
333
24k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
122
21k
Done Done
chrislema
186
16k
Transcript
Jingyuan Wang, Jiangtao Wen, Jun Zhang and Yuxing Han Tsinghua
National Laboratory for Information Science and Technology
Introduction Background and Motivation The TCP-FIT Algorithm
Experimental Results Conclusion
TCP ◦ The Transmission Control Protocol TCP Reno/
TCP New Reno ◦ Pros Reliable Sequential ◦ Cons Wireless network high Bandwidth Delay Product (BDP)
Recent development ◦ Wireless TCP Westwood TCP
Veno ◦ BDP TCP Compound TCP CUBIC FAST TCP TCP-FIT ◦ Inspired by parallel TCP
Introduction Background and Motivation The TCP-FIT Algorithm
Experimental Results Conclusion
What is a Congestion Algorithm? Classification ◦ Loss-based
◦ Delay-based ◦ Hybrid
Loss-Based ◦ TCP Reno, TCP CUBIC ◦ TCP BIC
, High Speed TCP Definition Assumption ◦ Constraints In wireless scenarios In BDP scenarios
Delay-Based ◦ TCP Vegas and FAST TCP Definition
Assumption ◦ Queuing delay = RTT – Propagation delay. ◦ Pros Resilient -> good ◦ Cons bandwidth starvation
Hybrid TCP ◦ Veno, TCP Westwood, TCP Illinois, H-TCP
and Fusion TCP ◦ Compound TCP Cons ◦ Poor in BDP-wireless-hybrid scenarios
TCP-FIT ◦ Parallel TCP ◦ E.g. GridFTP and E-MulTCP
◦ Pros: Utilization Good in wireless and BDP ◦ Cons: Compatibility Fairness
Introduction Background and Motivation The TCP-FIT Algorithm
Experimental Results Conclusion
Notation of parameter ◦ : Size of congestion window
� ◦ : RTT time ◦ : PLR, Packet Loss Rate ◦ : Throughput of the network ◦ : Propagation delay ◦ : Queuing delay
Object ◦ achieve N times throughput of the TCP
Reno ◦ meanwhile maintain fairness TCP Reno ◦ AIMD : ← + 1 : ← − 1 2
MulTCP ◦ : ← + ◦ : ← −
1 2 Proposed Method ◦ : ← + ◦ : ← − 2 3+1
Assumption: one loss adjustment is enough ◦ Based on
AIMD ◦ = − 2 3+1 ◦ + = 2� ◦ We get ◦ = 3+1 3 �
In a certain length of time( we set it
as k*RTT time ) ◦ ∆ = 1 − � + � 𝑘𝑘 � � 2 3+1 Then ◦ ̇ = ∆ ̇ = 𝑇𝑇 − 𝑋𝑋 � 2 3+1 � 3+1 3 � Let ̇ = 0 then we get = 32 2 � 2 ( = � 𝑇𝑇 ) ◦ = 1 𝑇𝑇 3 2 1
The is adjusted adaptively. Update of is ◦ 𝑖𝑖+1
= 1, 𝑖𝑖 + − −𝑡𝑡 𝑖𝑖 The equilibrium of is ◦ = � −𝑡𝑡 = 𝑇𝑇 ◦ (denote α as α = (𝑡𝑡 − 𝑡𝑡 )/𝑡𝑡 ) ◦ (Easy to know = = + = 𝑡𝑡 + ) ◦ = 2 3 1
Network Utilization ◦ = 1 𝑇𝑇 3 2 1
vs = 1 𝑇𝑇 3 2 1 Fairness ◦ RTT-fairness η = 𝑖𝑖 = 1 vs η = 𝑖𝑖 = + 𝑖𝑖+ ( = = + ) ◦ Inter-fairness = 𝑇𝑇∗−𝑇𝑇′ 𝑇𝑇∗
Introduction Background and Motivation The TCP-FIT Algorithm
Experimental Results Conclusion
BDP Scenarios
Wireless Scenarios
Inter-fairness
RTT-fairness
Cons: ◦ a low speed ADSL network with large
bandwidth variations ◦ Due to simplistic model of bandwidth estimation compared with FAST.
Thank you