TCP Incast avoidance based on connection serialization in data center networks

Shigeyuki Osada, Ryo Miyayama, Yukinobu Fukushima, Tokumi Yokohira

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

In distributed file systems, a well-known congestion collapse called TCP incast (Incast briefly) occurs because many servers almost simultaneously send data to the same client and then many packets overflow the port buffer of the link connecting to the client. Incast leads to throughput degradation in the network. In this paper, we propose three methods to avoid Incast based on the fact that the bandwidth-delay product is small in current data center networks. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second and third methods are methods which overlap the slow start period of the next connection with the current established connection to improve throughput in the first method. Numerical results from extensive simulation runs show the effectiveness of our three proposed methods.

Original languageEnglish
Pages (from-to)83-102
Number of pages20
JournalInternational Journal of Computer Networks and Communications
Volume8
Issue number4
DOIs
Publication statusPublished - Jan 1 2016

Fingerprint

Throughput
Servers
Bandwidth
Degradation

Keywords

  • Data Center
  • Distributed File System
  • TCP
  • TCP Incast

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Hardware and Architecture

Cite this

TCP Incast avoidance based on connection serialization in data center networks. / Osada, Shigeyuki; Miyayama, Ryo; Fukushima, Yukinobu; Yokohira, Tokumi.

In: International Journal of Computer Networks and Communications, Vol. 8, No. 4, 01.01.2016, p. 83-102.

Research output: Contribution to journalArticle

@article{d7c6fdbf797c4898b19dba5f0bca16ae,
title = "TCP Incast avoidance based on connection serialization in data center networks",
abstract = "In distributed file systems, a well-known congestion collapse called TCP incast (Incast briefly) occurs because many servers almost simultaneously send data to the same client and then many packets overflow the port buffer of the link connecting to the client. Incast leads to throughput degradation in the network. In this paper, we propose three methods to avoid Incast based on the fact that the bandwidth-delay product is small in current data center networks. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second and third methods are methods which overlap the slow start period of the next connection with the current established connection to improve throughput in the first method. Numerical results from extensive simulation runs show the effectiveness of our three proposed methods.",
keywords = "Data Center, Distributed File System, TCP, TCP Incast",
author = "Shigeyuki Osada and Ryo Miyayama and Yukinobu Fukushima and Tokumi Yokohira",
year = "2016",
month = "1",
day = "1",
doi = "10.5121/ijcnc.2016.8406",
language = "English",
volume = "8",
pages = "83--102",
journal = "International Journal of Computer Networks and Communications",
issn = "0975-2293",
publisher = "Academy and Industry Research Collaboration Center (AIRCC)",
number = "4",

}

TY - JOUR

T1 - TCP Incast avoidance based on connection serialization in data center networks

AU - Osada, Shigeyuki

AU - Miyayama, Ryo

AU - Fukushima, Yukinobu

AU - Yokohira, Tokumi

PY - 2016/1/1

Y1 - 2016/1/1

N2 - In distributed file systems, a well-known congestion collapse called TCP incast (Incast briefly) occurs because many servers almost simultaneously send data to the same client and then many packets overflow the port buffer of the link connecting to the client. Incast leads to throughput degradation in the network. In this paper, we propose three methods to avoid Incast based on the fact that the bandwidth-delay product is small in current data center networks. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second and third methods are methods which overlap the slow start period of the next connection with the current established connection to improve throughput in the first method. Numerical results from extensive simulation runs show the effectiveness of our three proposed methods.

AB - In distributed file systems, a well-known congestion collapse called TCP incast (Incast briefly) occurs because many servers almost simultaneously send data to the same client and then many packets overflow the port buffer of the link connecting to the client. Incast leads to throughput degradation in the network. In this paper, we propose three methods to avoid Incast based on the fact that the bandwidth-delay product is small in current data center networks. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second and third methods are methods which overlap the slow start period of the next connection with the current established connection to improve throughput in the first method. Numerical results from extensive simulation runs show the effectiveness of our three proposed methods.

KW - Data Center

KW - Distributed File System

KW - TCP

KW - TCP Incast

UR - http://www.scopus.com/inward/record.url?scp=85044287687&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044287687&partnerID=8YFLogxK

U2 - 10.5121/ijcnc.2016.8406

DO - 10.5121/ijcnc.2016.8406

M3 - Article

VL - 8

SP - 83

EP - 102

JO - International Journal of Computer Networks and Communications

JF - International Journal of Computer Networks and Communications

SN - 0975-2293

IS - 4

ER -