One document matched: draft-floyd-tcp-slowstart-01.ps
%!PS-Adobe-3.0
%%BoundingBox: 24 24 588 768
%%Title: Enscript Output
%%For: Sally Floyd
%%Creator: GNU enscript 1.6.1
%%CreationDate: Wed Aug 7 11:52:18 2002
%%Orientation: Portrait
%%Pages: (atend)
%%DocumentMedia: Letter 612 792 0 () ()
%%DocumentNeededResources: (atend)
%%EndComments
%%BeginProlog
%%BeginResource: procset Enscript-Prolog 1.6 1
%
% Procedures.
%
/_S { % save current state
/_s save def
} def
/_R { % restore from saved state
_s restore
} def
/S { % showpage protecting gstate
gsave
showpage
grestore
} bind def
/MF { % fontname newfontname -> - make a new encoded font
/newfontname exch def
/fontname exch def
/fontdict fontname findfont def
/newfont fontdict maxlength dict def
fontdict {
exch
dup /FID eq {
% skip FID pair
pop pop
} {
% copy to the new font dictionary
exch newfont 3 1 roll put
} ifelse
} forall
newfont /FontName newfontname put
% insert only valid encoding vectors
encoding_vector length 256 eq {
newfont /Encoding encoding_vector put
} if
newfontname newfont definefont pop
} def
/SF { % fontname width height -> - set a new font
/height exch def
/width exch def
findfont
[width 0 0 height 0 0] makefont setfont
} def
/SUF { % fontname width height -> - set a new user font
/height exch def
/width exch def
/F-gs-user-font MF
/F-gs-user-font width height SF
} def
/M {moveto} bind def
/s {show} bind def
/Box { % x y w h -> - define box path
/d_h exch def /d_w exch def /d_y exch def /d_x exch def
d_x d_y moveto
d_w 0 rlineto
0 d_h rlineto
d_w neg 0 rlineto
closepath
} def
/bgs { % x y height blskip gray str -> - show string with bg color
/str exch def
/gray exch def
/blskip exch def
/height exch def
/y exch def
/x exch def
gsave
x y blskip sub str stringwidth pop height Box
gray setgray
fill
grestore
x y M str s
} def
% Highlight bars.
/highlight_bars { % nlines lineheight output_y_margin gray -> -
gsave
setgray
/ymarg exch def
/lineheight exch def
/nlines exch def
% This 2 is just a magic number to sync highlight lines to text.
0 d_header_y ymarg sub 2 sub translate
/cw d_output_w cols div def
/nrows d_output_h ymarg 2 mul sub lineheight div cvi def
% for each column
0 1 cols 1 sub {
cw mul /xp exch def
% for each rows
0 1 nrows 1 sub {
/rn exch def
rn lineheight mul neg /yp exch def
rn nlines idiv 2 mod 0 eq {
% Draw highlight bar. 4 is just a magic indentation.
xp 4 add yp cw 8 sub lineheight neg Box fill
} if
} for
} for
grestore
} def
% Line highlight bar.
/line_highlight { % x y width height gray -> -
gsave
/gray exch def
Box gray setgray fill
grestore
} def
% Column separator lines.
/column_lines {
gsave
.1 setlinewidth
0 d_footer_h translate
/cw d_output_w cols div def
1 1 cols 1 sub {
cw mul 0 moveto
0 d_output_h rlineto stroke
} for
grestore
} def
% Column borders.
/column_borders {
gsave
.1 setlinewidth
0 d_footer_h moveto
0 d_output_h rlineto
d_output_w 0 rlineto
0 d_output_h neg rlineto
closepath stroke
grestore
} def
% Do the actual underlay drawing
/draw_underlay {
ul_style 0 eq {
ul_str true charpath stroke
} {
ul_str show
} ifelse
} def
% Underlay
/underlay { % - -> -
gsave
0 d_page_h translate
d_page_h neg d_page_w atan rotate
ul_gray setgray
ul_font setfont
/dw d_page_h dup mul d_page_w dup mul add sqrt def
ul_str stringwidth pop dw exch sub 2 div ul_h_ptsize -2 div moveto
draw_underlay
grestore
} def
/user_underlay { % - -> -
gsave
ul_x ul_y translate
ul_angle rotate
ul_gray setgray
ul_font setfont
0 0 ul_h_ptsize 2 div sub moveto
draw_underlay
grestore
} def
% Page prefeed
/page_prefeed { % bool -> -
statusdict /prefeed known {
statusdict exch /prefeed exch put
} {
pop
} ifelse
} def
% Wrapped line markers
/wrapped_line_mark { % x y charwith charheight type -> -
/type exch def
/h exch def
/w exch def
/y exch def
/x exch def
type 2 eq {
% Black boxes (like TeX does)
gsave
0 setlinewidth
x w 4 div add y M
0 h rlineto w 2 div 0 rlineto 0 h neg rlineto
closepath fill
grestore
} {
type 3 eq {
% Small arrows
gsave
.2 setlinewidth
x w 2 div add y h 2 div add M
w 4 div 0 rlineto
x w 4 div add y lineto stroke
x w 4 div add w 8 div add y h 4 div add M
x w 4 div add y lineto
w 4 div h 8 div rlineto stroke
grestore
} {
% do nothing
} ifelse
} ifelse
} def
% EPSF import.
/BeginEPSF {
/b4_Inc_state save def % Save state for cleanup
/dict_count countdictstack def % Count objects on dict stack
/op_count count 1 sub def % Count objects on operand stack
userdict begin
/showpage { } def
0 setgray 0 setlinecap
1 setlinewidth 0 setlinejoin
10 setmiterlimit [ ] 0 setdash newpath
/languagelevel where {
pop languagelevel
1 ne {
false setstrokeadjust false setoverprint
} if
} if
} bind def
/EndEPSF {
count op_count sub { pos } repeat % Clean up stacks
countdictstack dict_count sub { end } repeat
b4_Inc_state restore
} bind def
% Check PostScript language level.
/languagelevel where {
pop /gs_languagelevel languagelevel def
} {
/gs_languagelevel 1 def
} ifelse
%%EndResource
%%BeginResource: procset Enscript-Encoding-88591 1.6 1
/encoding_vector [
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/space /exclam /quotedbl /numbersign
/dollar /percent /ampersand /quoteright
/parenleft /parenright /asterisk /plus
/comma /hyphen /period /slash
/zero /one /two /three
/four /five /six /seven
/eight /nine /colon /semicolon
/less /equal /greater /question
/at /A /B /C
/D /E /F /G
/H /I /J /K
/L /M /N /O
/P /Q /R /S
/T /U /V /W
/X /Y /Z /bracketleft
/backslash /bracketright /asciicircum /underscore
/quoteleft /a /b /c
/d /e /f /g
/h /i /j /k
/l /m /n /o
/p /q /r /s
/t /u /v /w
/x /y /z /braceleft
/bar /braceright /tilde /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef
/space /exclamdown /cent /sterling
/currency /yen /brokenbar /section
/dieresis /copyright /ordfeminine /guillemotleft
/logicalnot /hyphen /registered /macron
/degree /plusminus /twosuperior /threesuperior
/acute /mu /paragraph /bullet
/cedilla /onesuperior /ordmasculine /guillemotright
/onequarter /onehalf /threequarters /questiondown
/Agrave /Aacute /Acircumflex /Atilde
/Adieresis /Aring /AE /Ccedilla
/Egrave /Eacute /Ecircumflex /Edieresis
/Igrave /Iacute /Icircumflex /Idieresis
/Eth /Ntilde /Ograve /Oacute
/Ocircumflex /Otilde /Odieresis /multiply
/Oslash /Ugrave /Uacute /Ucircumflex
/Udieresis /Yacute /Thorn /germandbls
/agrave /aacute /acircumflex /atilde
/adieresis /aring /ae /ccedilla
/egrave /eacute /ecircumflex /edieresis
/igrave /iacute /icircumflex /idieresis
/eth /ntilde /ograve /oacute
/ocircumflex /otilde /odieresis /divide
/oslash /ugrave /uacute /ucircumflex
/udieresis /yacute /thorn /ydieresis
] def
%%EndResource
%%EndProlog
%%BeginSetup
%%IncludeResource: font Courier-Bold
%%IncludeResource: font Courier
/HFpt_w 10 def
/HFpt_h 10 def
/Courier-Bold /HF-gs-font MF
/HF /HF-gs-font findfont [HFpt_w 0 0 HFpt_h 0 0] makefont def
/Courier /F-gs-font MF
/F-gs-font 10 10 SF
/#copies 1 def
% Pagedevice definitions:
gs_languagelevel 1 gt {
<<
/PageSize [612 792]
>> setpagedevice
} if
/d_page_w 564 def
/d_page_h 744 def
/d_header_x 0 def
/d_header_y 744 def
/d_header_w 564 def
/d_header_h 0 def
/d_footer_x 0 def
/d_footer_y 0 def
/d_footer_w 564 def
/d_footer_h 0 def
/d_output_w 564 def
/d_output_h 744 def
/cols 1 def
%%EndSetup
%%Page: (1) 1
%%BeginPageSetup
_S
24 24 translate
/pagenum 1 def
/fname (draft-floyd-tcp-slowstart-01.txt) def
/fdir () def
/ftail (draft-floyd-tcp-slowstart-01.txt) def
/user_header_p false def
%%EndPageSetup
5 665 M
(Internet Engineering Task Force Sally Floyd) s
5 654 M
(INTERNET DRAFT ICSI) s
5 643 M
(draft-floyd-tcp-slowstart-01.txt August, 2002) s
5 599 M
( Limited Slow-Start for TCP with Large Congestion Windows) s
5 555 M
( Status of this Memo) s
5 522 M
( This document is an Internet-Draft and is in full conformance with) s
5 511 M
( all provisions of Section 10 of RFC2026.) s
5 489 M
( Internet-Drafts are working documents of the Internet Engineering) s
5 478 M
( Task Force \(IETF\), its areas, and its working groups. Note that) s
5 467 M
( other groups may also distribute working documents as Internet-) s
5 456 M
( Drafts.) s
5 434 M
( Internet-Drafts are draft documents valid for a maximum of six months) s
5 423 M
( and may be updated, replaced, or obsoleted by other documents at any) s
5 412 M
( time. It is inappropriate to use Internet- Drafts as reference) s
5 401 M
( material or to cite them other than as "work in progress.") s
5 379 M
( The list of current Internet-Drafts can be accessed at) s
5 368 M
( http://www.ietf.org/ietf/1id-abstracts.txt) s
5 346 M
( The list of Internet-Draft Shadow Directories can be accessed at) s
5 335 M
( http://www.ietf.org/shadow.html.) s
5 313 M
(Abstract) s
5 291 M
( This note proposes a modification for TCP's slow-start for use with) s
5 280 M
( TCP connections with large congestion windows. For TCP connections) s
5 269 M
( that are able to use congestion windows of thousands \(or tens of) s
5 258 M
( thousands\) of MSS-sized segments \(for MSS the sender's MAXIMUM) s
5 247 M
( SEGMENT SIZE\), the current slow-start procedure can result in) s
5 236 M
( increasing the congestion window by thousands of segments in a single) s
5 225 M
( round-trip time. Such an increase can easily result in thousands of) s
5 214 M
( packets being dropped in one round-trip time. This is often counter-) s
5 203 M
( productive for the TCP flow itself, and is also hard on the rest of) s
5 192 M
( the traffic sharing the congested link. This note proposes Limited) s
5 181 M
( Slow-Start as an optional mechanism for limiting the number of) s
5 170 M
( segments by which the congestion window is increased for one window) s
5 159 M
( of data during slow-start, in order to improve performance for TCP) s
5 148 M
( connections with large congestion windows.) s
5 104 M
(Floyd Experimental [Page 1]) s
_R
S
%%Page: (2) 2
%%BeginPageSetup
_S
24 24 translate
/pagenum 2 def
/fname (draft-floyd-tcp-slowstart-01.txt) def
/fdir () def
/ftail (draft-floyd-tcp-slowstart-01.txt) def
/user_header_p false def
%%EndPageSetup
5 698 M
(draft-floyd-tcp-slowstart-01 August 2002) s
5 665 M
( Changes from draft-floyd-tcp-slowstart-00.txt:) s
5 643 M
( * Small changes in presentation.) s
5 621 M
( * The addition of a section of Experiments.) s
5 599 M
( * More citations to related work.) s
5 577 M
(1. Introduction) s
5 555 M
( This note proposes a modification for TCP's slow-start for use with) s
5 544 M
( TCP connections with large congestion windows. For TCP connections) s
5 533 M
( that are able to use congestion windows of thousands \(or tens of) s
5 522 M
( thousands\) of MSS-sized segments \(for MSS the sender's MAXIMUM) s
5 511 M
( SEGMENT SIZE\), the current slow-start procedure can result in) s
5 500 M
( increasing the congestion window by thousands of segments in a single) s
5 489 M
( round-trip time. Such an increase can easily result in thousands of) s
5 478 M
( packets being dropped in one round-trip time. This is often counter-) s
5 467 M
( productive for the TCP flow itself, and is also hard on the rest of) s
5 456 M
( the traffic sharing the congested link. This note proposes Limited) s
5 445 M
( Slow-Start, limiting the number of segments by which the congestion) s
5 434 M
( window is increased for one window of data during slow-start, in) s
5 423 M
( order to improve performance for TCP connections with large) s
5 412 M
( congestion windows.) s
5 390 M
( When slow-start results in a large increase in the congestion window) s
5 379 M
( in one round-trip time, a large number of packets might be dropped in) s
5 368 M
( the network \(even with carefully-tuned active queue management) s
5 357 M
( mechanisms in the routers\). This drop of a large number of packets) s
5 346 M
( in the network can result in unnecessary retransmit timeouts for the) s
5 335 M
( TCP connection. The TCP connection could end up in congestion) s
5 324 M
( avoidance phase with a very small congestion window, and could take a) s
5 313 M
( large number of round-trip times to recover its old congestion) s
5 302 M
( window. This poor performance is illustrated in [F02].) s
5 280 M
(2. The Proposal for Limited Slow-Start) s
5 258 M
( Limited Slow-Start introduces a parameter, "max_ssthresh", and the) s
5 247 M
( slow-start is only modified for values of the congestion window) s
5 236 M
( "cwnd" greater than "max_ssthresh". That is, during Slow-Start, when) s
5 214 M
( cwnd <= max_ssthresh,) s
5 192 M
( cwnd is increased by one MSS \(MAXIMUM SEGMENT SIZE\) for every) s
5 181 M
( arriving ACK \(acknowledgement\) during slow-start, as is always the) s
5 170 M
( case. During Limited Slow-Start, when) s
5 104 M
(Floyd Experimental [Page 2]) s
_R
S
%%Page: (3) 3
%%BeginPageSetup
_S
24 24 translate
/pagenum 3 def
/fname (draft-floyd-tcp-slowstart-01.txt) def
/fdir () def
/ftail (draft-floyd-tcp-slowstart-01.txt) def
/user_header_p false def
%%EndPageSetup
5 698 M
(draft-floyd-tcp-slowstart-01 August 2002) s
5 665 M
( max_ssthresh < cwnd <= ssthresh,) s
5 643 M
( the invariant is maintained that the congestion window is increased) s
5 632 M
( during slow-start by at most max_ssthresh/2 MSS per round-trip time.) s
5 621 M
( This is done as follows:) s
5 599 M
( For each arriving ACK in slow-start:) s
5 588 M
( If \(cwnd <= max_ssthresh\)) s
5 577 M
( cwnd += MSS;) s
5 566 M
( else) s
5 555 M
( K = int\(cwnd/\(0.5 max_ssthresh\)\);) s
5 544 M
( cwnd += int\(MSS/K\);) s
5 522 M
( Thus during Limited Slow-Start the window is increased by 1/K MSS for) s
5 511 M
( each arriving ACK, for K = int\(cwnd/\(0.5 max_ssthresh\)\), instead of) s
5 500 M
( by 1 MSS as in the standard slow-start [RFC2581].) s
5 478 M
( When) s
5 456 M
( ssthresh < cwnd,) s
5 434 M
( slow-start is exited, and the sender is in the Congestion Avoidance) s
5 423 M
( phase.) s
5 401 M
( Our current recommendation would be for max_ssthresh to be set to 100) s
5 390 M
( MSS. \(This is illustrated in the NS simulator, for snapshots after) s
5 379 M
( May 1, 2002, in the tests "./test-all-tcpHighspeed tcp1A" and) s
5 368 M
( "./test-all-tcpHighspeed tcpHighspeed1" in the subdirectory) s
5 357 M
( "tcl/lib".\) However, max_ssthresh is an optional, configurable) s
5 346 M
( parameter. TCP implementations that do not wish to use Limited Slow-) s
5 335 M
( Start would set max_ssthresh to Infinity.) s
5 313 M
( With Limited Slow-Start, when the congestion window is greater than) s
5 302 M
( max_ssthresh the window is increased by at most 1/2 MSS for each) s
5 291 M
( arriving ACK, when the congestion window is greater than 1.5) s
5 280 M
( max_ssthresh the window is increased by at most 1/3 MSS for each) s
5 269 M
( arriving ACK, and so on.) s
5 247 M
( With Limited Slow-Start it takes:) s
5 225 M
( log\(max_ssthresh\)) s
5 203 M
( round-trip times to reach a congestion window of max_ssthresh, and it) s
5 192 M
( takes:) s
5 170 M
( log\(max_ssthresh\) + \(cwnd - max_ssthresh\)/\(max_ssthresh/2\)) s
5 148 M
( round-trip times to reach a congestion window of cwnd, for a) s
5 104 M
(Floyd Experimental [Page 3]) s
_R
S
%%Page: (4) 4
%%BeginPageSetup
_S
24 24 translate
/pagenum 4 def
/fname (draft-floyd-tcp-slowstart-01.txt) def
/fdir () def
/ftail (draft-floyd-tcp-slowstart-01.txt) def
/user_header_p false def
%%EndPageSetup
5 698 M
(draft-floyd-tcp-slowstart-01 August 2002) s
5 665 M
( congestion window greater than max_ssthresh.) s
5 643 M
( Thus, with Limited Slow-Start with max_ssthresh set to 100 MSS, it) s
5 632 M
( would take 836 round-trip times to reach a congestion window of) s
5 621 M
( 83,000 packets, compared to 16 round-trip times without Limited Slow-) s
5 610 M
( Start \(assuming no packet drops\). With Limited Slow-Start, the) s
5 599 M
( largest transient queue during slow-start would be 100 packets;) s
5 588 M
( without Limited Slow-Start, the transient queue during Slow-Start) s
5 577 M
( would reach more than 32,000 packets.) s
5 555 M
( By limiting the maximum increase in the congestion window in a round-) s
5 544 M
( trip time, Limited Slow-Start can reduce the number of drops during) s
5 533 M
( slow-start, and improve the performance of TCP connections with large) s
5 522 M
( congestion windows.) s
5 500 M
(3. Experimental Results) s
5 478 M
( Tom Dunigan has added Limited Slow-Start to the Linux 2.4.16 Web100) s
5 467 M
( kernel, and performed experiments comparing TCP with and without) s
5 456 M
( Limited Slow-Start [D02]. Preliminary results show improved) s
5 445 M
( performance for TCPs using Limited Slow-Start. There are also) s
5 434 M
( several experiments comparing different values for max_ssthresh.) s
5 412 M
( More experiments will be needed to more fully evaluate Limited Slow-) s
5 401 M
( Start.) s
5 379 M
(4. Related Proposals) s
5 357 M
( There has been considerable research on mechanisms for the TCP sender) s
5 346 M
( to learn about the limitations of the available bandwidth, and to) s
5 335 M
( exit slow-start before receiving a congestion indication from the) s
5 324 M
( network [VEGAS,H96]. Other proposals set TCP's slow-start parameter) s
5 313 M
( ssthresh based on information from previous TCP connections to the) s
5 302 M
( same destination [WS95,G00]. This draft proposes a simple limitation) s
5 291 M
( on slow-start that can be effective in some cases even in the absence) s
5 280 M
( of such mechanisms. The max_ssthresh parameter does not replace) s
5 269 M
( ssthresh, but is an additional parameter. Thus, Limited Slow-Start) s
5 258 M
( could be used in addition to mechanisms for setting ssthresh.) s
5 236 M
( Rate-based pacing has also been proposed to improve the performance) s
5 225 M
( of TCP during slow-start [VH97,AD98,KCRP99,ASA00]. We believe that) s
5 214 M
( rate-based pacing could be of significant benefit, and could be used) s
5 203 M
( in addition to the Limited Slow-Start in this proposal.) s
5 181 M
( Appropriate Byte Counting [A01] proposes that TCP increase its) s
5 170 M
( congestion window as a function of the number of bytes acknowledged,) s
5 159 M
( rather than as a function of the number of ACKs received.) s
5 148 M
( Appropriate Byte Counting is largely orthogonal to this proposal for) s
5 104 M
(Floyd Experimental [Page 4]) s
_R
S
%%Page: (5) 5
%%BeginPageSetup
_S
24 24 translate
/pagenum 5 def
/fname (draft-floyd-tcp-slowstart-01.txt) def
/fdir () def
/ftail (draft-floyd-tcp-slowstart-01.txt) def
/user_header_p false def
%%EndPageSetup
5 698 M
(draft-floyd-tcp-slowstart-01 August 2002) s
5 665 M
( Limited Slow-Start.) s
5 643 M
( Limited Slow-Start is also orthogonal to other proposals to change) s
5 632 M
( mechanisms for exiting slow-start. For example, FACK TCP includes an) s
5 621 M
( overdamping mechanism to decrease the congestion window somewhat more) s
5 610 M
( aggressively when a loss occurs during slow-start [MM96]. It is also) s
5 599 M
( true that larger values for the MSS would reduce the size of the) s
5 588 M
( congestion window in units of MSS needed to fill a given pipe, and) s
5 577 M
( therefore would reduce the size of the transient queue in units of) s
5 566 M
( MSS.) s
5 544 M
(5. Acknowledgements) s
5 522 M
( This proposal is part of a larger proposal for HighSpeed TCP for TCP) s
5 511 M
( connections with large congestion windows, and resulted from) s
5 500 M
( simulations done by Evandro de Souza, in joint work with Deb Agarwal.) s
5 489 M
( This proposal for Limited Slow-Start drew in part from discussions) s
5 478 M
( with Tom Kelly, who has used a similar modified slow-start in his own) s
5 467 M
( research with congestion control for high-bandwidth connections. We) s
5 456 M
( also thank Tom Dunigan for his experiments with Limited Slow-Start.) s
5 434 M
( We thank Andrei Gurtov, Reiner Ludwig, members of the End-to-End) s
5 423 M
( Research Group, and members of the Transport Area Working Group, for) s
5 412 M
( feedback on this document.) s
5 390 M
(6. Normative References) s
5 368 M
( [A01] Mark Allman, "TCP Congestion Control with Appropriate Byte) s
5 357 M
( Counting", draft-allman-tcp-abc-02.txt, internet-draft, work-in-) s
5 346 M
( progress, November 2001.) s
5 324 M
( [RFC2581] M. Allman and V. Paxson, "TCP Congestion Control", RFC) s
5 313 M
( 2581, April 1999.) s
5 291 M
(7. Informative References) s
5 269 M
( [AD98] Mohit Aron and Peter Druschel, "TCP: Improving Start-up) s
5 258 M
( Dynamics by Adaptive Timers and Congestion Control"", TR98-318, Rice) s
5 247 M
( University, 1998. URL "http://cs-) s
5 236 M
( tr.cs.rice.edu/Dienst/UI/2.0/Describe/ncstrl.rice_cs/TR98-318/".) s
5 214 M
( [ASA00] A. Aggarwal, S. Savage, and T. Anderson, "Understanding the) s
5 203 M
( Performance of TCP Pacing", Proceedings of the 2000 IEEE Infocom) s
5 192 M
( Conference, Tel-Aviv, Israel, March, 2000. URL) s
5 181 M
( "http://www.cs.ucsd.edu/~savage/".) s
5 159 M
( [D02] T. Dunigan, "Floyd's TCP slow-start and AIMD mods", 2002. URL) s
5 148 M
( "http://www.csm.ornl.gov/~dunigan/net100/floyd.html".) s
5 104 M
(Floyd Experimental [Page 5]) s
_R
S
%%Page: (6) 6
%%BeginPageSetup
_S
24 24 translate
/pagenum 6 def
/fname (draft-floyd-tcp-slowstart-01.txt) def
/fdir () def
/ftail (draft-floyd-tcp-slowstart-01.txt) def
/user_header_p false def
%%EndPageSetup
5 698 M
(draft-floyd-tcp-slowstart-01 August 2002) s
5 665 M
( [F02] S. Floyd, "Performance Problems with TCP's Slow-Start", 2002.) s
5 654 M
( URL "http://www.icir.org/floyd/hstcp/slowstart/".) s
5 632 M
( [G00] A. Gurtov, "TCP Performance in the Presence of Congestion and) s
5 621 M
( Corruption Losses", Master's Thesis, University of Helsinki,) s
5 610 M
( Department of Computer Science, Helsinki, December 2000. URL) s
5 599 M
( "http://www.cs.helsinki.fi/u/gurtov/papers/ms_thesis.html".) s
5 577 M
( [H96] J. C. Hoe, "Improving the Start-up Behavior of a Congestion) s
5 566 M
( Control Scheme for TCP", SIGCOMM 96, 1996. URL) s
5 555 M
( "http://www.acm.org/sigcomm/sigcomm96/program.html".) s
5 533 M
( [KCRP99] J. Kulik, R. Coulter, D. Rockwell, and C. Partridge, "A) s
5 522 M
( Simulation Study of Paced TCP", BBN Technical Memorandum No. 1218,) s
5 511 M
( 1999. URL "http://mimas.lcs.mit.edu/~jokulik/tcppacing.html".) s
5 489 M
( [MM96] M. Mathis and J. Mahdavi, "Forward Acknowledgment: Refining) s
5 478 M
( TCP Congestion Control", SIGCOMM, August 1996.) s
5 456 M
( [VEGAS] Vegas Web Page, University of Arizona. URL) s
5 445 M
( "http://www.cs.arizona.edu/protocols/".) s
5 423 M
( [VH97] Vikram Visweswaraiah and John Heidemann, "Rate Based Pacing) s
5 412 M
( for TCP", 1997. URL) s
5 401 M
( "http://www.isi.edu/lsam/publications/rate_based_pacing/".) s
5 379 M
( [WS95] G. Wright and W. Stevens, "TCP/IP Illustrated", Volume 2,) s
5 368 M
( Addison-Wesley Publishing Company, 1995.) s
5 346 M
(8. Security Considerations) s
5 324 M
( This proposal makes no changes to the underlying security of TCP.) s
5 302 M
(8. IANA Considerations) s
5 280 M
( There are no IANA considerations regarding this document.) s
5 258 M
( AUTHORS' ADDRESSES) s
5 225 M
( Sally Floyd) s
5 214 M
( Phone: +1 \(510\) 666-2989) s
5 203 M
( ICIR \(ICSI Center for Internet Research\)) s
5 192 M
( Email: floyd@icir.org) s
5 181 M
( URL: http://www.icir.org/floyd/) s
5 159 M
( This draft was created in August 2002.) s
5 104 M
(Floyd Experimental [Page 6]) s
_R
S
%%Trailer
%%Pages: 6
%%DocumentNeededResources: font Courier-Bold Courier
%%EOF
| PAFTECH AB 2003-2026 | 2026-04-22 14:08:33 |