Internet Engineering Task Force Yunhong Gu
Internet Draft Robert L. Grossman
Document: draft-gg-udt-01.txt University of Illinois at Chicago
Expires: February 2005 August 2004
UDT: A Transport Protocol for Data Intensive Applications
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC 2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that other
groups may also distribute working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Abstract
This document proposes a new UDP based Data Transfer protocol, also
known as UDT. UDT can be used in high bandwidth-delay product (BDP)
networks to utilize the abundant network resource efficiently and
fairly. It is expected to be used in distributed data intensive
applications over high-speed wide area networks, such as scientific
computing on computational grids.
The current most widely used TCP version (TCP NewReno/SACK) does not
work well in such environments due to its slow discovery and recovery
to the available bandwidth as the network BDP increases. In addition,
it exhibits "unfairness" when concurrent TCP flows have different
round trip time (RTT).
UDT is an application layer solution by introducing new congestion
control and reliability control. The protocol if defined upon UDP.
The congestion control combines both rate-based and window-based
methods and uses bandwidth estimation techniques to dynamically
configure the control parameters.
Gu, Grossman Expires - February 2005 [Page 1]
UDT August 2004
Conventions used in this document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119.
Table of Contents
1. Introduction...................................................2
2. Target Use Scenarios and Design Goals..........................3
3. Protocol Specification.........................................4
3.1 Overview...................................................4
3.2 Packet Structures..........................................4
3.3 Timing.....................................................6
3.4 The Sender's Algorithm.....................................7
3.5 The Receiver's Algorithm...................................8
3.6 Rate Control Algorithm....................................11
3.7 Flow Control Algorithm....................................12
3.8 Connection Setup and shutdown.............................12
3.9 Loss Information Compression Scheme.......................13
4. Efficiency and Fairness Characters............................14
Security Considerations..........................................14
Normative References.............................................15
Informative References...........................................15
Acknowledgments..................................................15
Author's Addresses...............................................16
1. Introduction
As the network bandwidth-delay product (BDP) increases, conventional
TCP implementations become inefficient. This is because its AIMD
(additive increase multiplicative decrease) algorithm reduces the TCP
congestion window drastically but fails to recover it to the
available bandwidth quickly. Theoretical flow level analysis has
shown that TCP becomes more vulnerable to packet loss as the BDP
increases higher [LM97].
Additionally, the unfairness of RTT bias inherent in TCP congestion
control also becomes more serious in distributed data intensive
applications. Concurrent TCP flows with different RTT will tend to
unfairly share the available bandwidth. Although small BDP networks
will share bandwidth relatively fairly using conventional TCP
implementations, networks with large BDP tend to suffer severe
unfairness issues under conventional TCP. This RTT bias severely
limits the effectiveness of distributed computations based on high-
speed wide-area networks as, for example, Grid-based computations in
Gu, Grossman Expires - February 2005 [Page 2]
UDT August 2004
an Internet2-based context.
As of today, improvements on the standard TCP still fails to satisfy
the efficiency and fairness (especially the RTT bias problem)
objectives in high BDP environments. Such TCP modifications as
presented in RFC 1423 (high-performance extensions), RFC 2018 (SACK),
RFC 2582 (New Reno), RFC 2883 (D-SACK), and RFC 2988 (RTO
calculation) do improve the efficiency more or less, but the
fundamental problem of the TCP AIMD algorithm is unsolved. HS TCP
(RFC 3649) achieved high utilization of bandwidth in high BDP
networks by radically changing the TCP congestion control algorithm,
but the fairness problem still remains.
Considering the background above, new transport protocol is needed to
support the high performance bulk data transfer in high BDP networks.
We propose a new application level transport protocol, named UDT, or
UDP based Data Transfer Protocol, together with a new congestion
control algorithm.
This document presents the two orthogonal parts of the UDT protocol
and the UDT congestion control algorithm. The application level
protocol, layering over UDP, can use other congestion control
algorithms; whereas the algorithm defined in this document can be
implemented in other protocols, such as TCP.
An example implementation of the protocol can be found at [UDT].
Detailed performance analysis of the congestion control algorithm can
be found in [GHG04].
2. Target Use Scenarios and Design Goals
UDT is expected to be used in the situations where a small number of
bulk sources share abundant bandwidth. Typical use scenario is the
computational grids [FKT01] built on wide area optical networks. A
few institutes run their data intensive distributed applications on
the networks, such as remote access to scientific instruments,
distributed data mining, and high-resolution media streams.
The main objectives of UDT are efficiency, fairness, and stability.
Single, or small number of, UDT flows should utilize all the
available bandwidth provided by the high-speed links, even if the
bandwidth changes drastically. Meanwhile, all concurrent flows must
share the bandwidth fairly, independent of the different bottleneck
bandwidth, start time, or RTT. Stability requires that the packet
sending rate should always converge to the available bandwidth very
quickly, and congestion collapse must be avoided.
UDT is NOT used to replace TCP in the Internet where the bottleneck
Gu, Grossman Expires - February 2005
没有合适的资源?快使用搜索试试~ 我知道了~
udt.sdk.2.3.tar.gz_TCP NAT_UDP 传输_udp 可靠传输_udt_传输 udp
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 146 浏览量
2022-09-21
03:35:04
上传
评论
收藏 154KB GZ 举报
温馨提示
共92个文件
htm:41个
gif:14个
cpp:13个
UDP的可靠传输项目UDT(开源),采用了类似TCP的华东窗口协议
资源推荐
资源详情
资源评论
收起资源包目录
udt.sdk.2.3.tar.gz (92个子文件)
udt2
LICENSE 26KB
Makefile.osx 34B
app
cc.h 18KB
sendfile.cpp 2KB
appclient.cpp 3KB
recvfile.cpp 2KB
Makefile 897B
appserver.cpp 3KB
src
list.cpp 22KB
udt.h 66KB
api.cpp 33KB
packet.cpp 10KB
udt.cpp 80KB
channel.cpp 13KB
common.cpp 22KB
buffer.cpp 20KB
Makefile 1016B
socket.cpp 9KB
README 2KB
Makefile.unix 1KB
sim
udt.cc 34KB
parser.cpp 3KB
udt.h 7KB
proc.tcl 2KB
test.tcl 752B
sim.tcl 3KB
COPYRIGHT 1001B
doc
main.htm 20KB
draft-gg-udt-01.txt 39KB
index.htm 2KB
hlp
ix_listp.gif 92B
ix_end.gif 63B
ix_up.gif 125B
ix_down.gif 123B
ix_link.gif 134B
ix_listm.gif 89B
ix_book.gif 122B
ix_space.gif 54B
ix_endp.gif 91B
ix_line.gif 63B
Thumbs.db 20KB
ix_list.gif 65B
ix_endm.gif 88B
ix_leaf.gif 118B
ix_open.gif 125B
doc
t-file.htm 4KB
reference.htm 986B
select.htm 7KB
error.htm 2KB
close.htm 4KB
intro.htm 3KB
function.htm 7KB
make.htm 5KB
trace.htm 5KB
socket.htm 3KB
structure.htm 17KB
listen.htm 4KB
tutorial.htm 2KB
t-cc.htm 6KB
t-firewall.htm 3KB
copy.htm 3KB
t-intro.htm 4KB
recvfile.htm 9KB
accept.htm 5KB
ccc.htm 9KB
treeview.css 326B
footer.htm 573B
sim.htm 5KB
bind.htm 4KB
peername.htm 4KB
opt.htm 15KB
send.htm 7KB
shutdown.htm 4KB
ecode.htm 9KB
t-data.htm 5KB
t-error.htm 4KB
faq.htm 11KB
header.htm 287B
connect.htm 5KB
overlap.htm 5KB
t-hello.htm 5KB
sendfile.htm 5KB
t-config.htm 5KB
recv.htm 8KB
sockname.htm 4KB
win
appclient.vcproj 3KB
appserver.vcproj 3KB
recvfile.vcproj 3KB
udt.vcproj 4KB
udt.sln 3KB
sendfile.vcproj 3KB
Makefile 179B
共 92 条
- 1
资源评论
钱亚锋
- 粉丝: 85
- 资源: 1万+
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功