One document matched: draft-irtf-tmrg-metrics-03.ps
%!PS-Adobe-3.0
%%BoundingBox: 24 24 588 768
%%Title: Enscript Output
%%For: Sally Floyd
%%Creator: GNU enscript 1.6.1
%%CreationDate: Sun Jun 25 21:10:51 2006
%%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 (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 731 M
(Internet Engineering Task Force Sally Floyd) s
5 720 M
(INTERNET-DRAFT Editor) s
5 709 M
(draft-irtf-tmrg-metrics-03.txt 24 June 2006) s
5 698 M
(Expires: December 2006) s
5 665 M
( Metrics for the Evaluation of Congestion Control Mechanisms) s
5 610 M
(Status of this Memo) s
5 588 M
( By submitting this Internet-Draft, each author represents that any) s
5 577 M
( applicable patent or other IPR claims of which he or she is aware) s
5 566 M
( have been or will be disclosed, and any of which he or she becomes) s
5 555 M
( aware will be disclosed, in accordance with Section 6 of BCP 79.) s
5 533 M
( Internet-Drafts are working documents of the Internet Engineering) s
5 522 M
( Task Force \(IETF\), its areas, and its working groups. Note that) s
5 511 M
( other groups may also distribute working documents as Internet-) s
5 500 M
( Drafts.) s
5 478 M
( Internet-Drafts are draft documents valid for a maximum of six) s
5 467 M
( months and may be updated, replaced, or obsoleted by other documents) s
5 456 M
( at any time. It is inappropriate to use Internet-Drafts as) s
5 445 M
( reference material or to cite them other than as "work in progress.") s
5 423 M
( The list of current Internet-Drafts can be accessed at) s
5 412 M
( http://www.ietf.org/ietf/1id-abstracts.txt.) s
5 390 M
( The list of Internet-Draft Shadow Directories can be accessed at) s
5 379 M
( http://www.ietf.org/shadow.html.) s
5 357 M
( This Internet-Draft will expire on December 2006.) s
5 335 M
(Abstract) s
5 313 M
( This document discusses the metrics to be considered in an) s
5 302 M
( evaluation of new or modified congestion control mechanisms for the) s
5 291 M
( Internet. This includes metrics for the evaluation of new transport) s
5 280 M
( protocols, of proposed modifications to TCP, of application-level) s
5 269 M
( congestion control, and of Active Queue Management \(AQM\) mechanisms) s
5 258 M
( in the router. This document is intended to be the first in a) s
5 247 M
( series of documents aimed at improving the models that we use in the) s
5 203 M
(Floyd [Page 1]) s
_R
S
%%Page: (2) 2
%%BeginPageSetup
_S
24 24 translate
/pagenum 2 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( evaluation of transport protocols.) s
5 665 M
( This document is a product of the Transport Modeling Research Group) s
5 654 M
( \(TRMG\), and has received detailed feedback from several members of) s
5 643 M
( the Research Group \(RG\). We are not aware of any controversies) s
5 632 M
( regarding the content of this document.) s
5 126 M
(Floyd [Page 2]) s
_R
S
%%Page: (3) 3
%%BeginPageSetup
_S
24 24 translate
/pagenum 3 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( Table of Contents) s
5 665 M
( 1. Conventions . . . . . . . . . . . . . . . . . . . . . . . . . 4) s
5 654 M
( 2. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . 5) s
5 643 M
( 3. Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . 6) s
5 632 M
( 3.1. Throughput, Delay, and Loss Rates. . . . . . . . . . . . 7) s
5 621 M
( 3.1.1. Throughput. . . . . . . . . . . . . . . . . . . . . 7) s
5 610 M
( 3.1.2. Delay . . . . . . . . . . . . . . . . . . . . . . . 8) s
5 599 M
( 3.1.3. Packet Loss Rates . . . . . . . . . . . . . . . . . 8) s
5 588 M
( 3.2. Response Times and Minimizing Oscillations . . . . . . . 9) s
5 577 M
( 3.2.1. Response to Changes . . . . . . . . . . . . . . . . 9) s
5 566 M
( 3.2.2. Minimizing Oscillations . . . . . . . . . . . . . . 10) s
5 555 M
( 3.3. Fairness and Convergence . . . . . . . . . . . . . . . . 10) s
5 544 M
( 3.4. Robustness for Challenging Environments. . . . . . . . . 12) s
5 533 M
( 3.5. Robustness to Failures and to Misbehaving) s
5 522 M
( Users . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13) s
5 511 M
( 3.6. Deployability. . . . . . . . . . . . . . . . . . . . . . 13) s
5 500 M
( 3.7. Metrics for Specific Types of Transport. . . . . . . . . 14) s
5 489 M
( 3.8. User-Based Metrics . . . . . . . . . . . . . . . . . . . 14) s
5 478 M
( 4. Metrics in the IP Performance Metrics \(IPPM\) Working) s
5 467 M
( Group. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14) s
5 456 M
( 5. Comments on Methodology . . . . . . . . . . . . . . . . . . . 14) s
5 445 M
( 6. Security Considerations . . . . . . . . . . . . . . . . . . . 15) s
5 434 M
( 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15) s
5 423 M
( 8. Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . 15) s
5 412 M
( Informative References . . . . . . . . . . . . . . . . . . . . . 15) s
5 401 M
( Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 18) s
5 390 M
( Full Copyright Statement . . . . . . . . . . . . . . . . . . . . 18) s
5 379 M
( Intellectual Property. . . . . . . . . . . . . . . . . . . . . . 18) s
5 126 M
(Floyd [Page 3]) s
_R
S
%%Page: (4) 4
%%BeginPageSetup
_S
24 24 translate
/pagenum 4 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
(1. Conventions) s
5 665 M
( The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",) s
5 654 M
( "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this) s
5 643 M
( document are to be interpreted as described in [RFC 2119]. TO BE) s
5 632 M
( DELETED BY THE RFC EDITOR UPON PUBLICATION:) s
5 610 M
( Changes from draft-irtf-tmrg-metrics-02.txt: * Added a few sentences) s
5 599 M
( to the Abstract about the) s
5 588 M
( status of the document.) s
5 566 M
( Changes from draft-irtf-tmrg-metrics-01.txt: * Added a discussion) s
5 555 M
( about the metrics in IPPM.) s
5 533 M
( Changes from draft-irtf-tmrg-metrics-01c.txt:) s
5 511 M
( * Added to the discussion of network-based, flow-based,) s
5 500 M
( and user-based metrics, based on email from Dado Colussi,) s
5 489 M
( Sean Moore, Damon Wischik, Dah Ming Chiu, and others.) s
5 467 M
( * Changed "packet drop rate" to "packet loss rate".) s
5 456 M
( Suggestion from Nelson Fonseca.) s
5 434 M
( * Added a discussion of the Colussi et al. paper on a new) s
5 423 M
( definition of fairness.) s
5 401 M
( * Added a discussion of the Chiu and Tan paper on redefining) s
5 390 M
( fairness for inelastic traffic.) s
5 368 M
( Changes from draft-irtf-tmrg-metrics-01b.txt:) s
5 346 M
( * Added a discussion of goodput vs. throughput.) s
5 335 M
( Suggestion from Nelson Fonseca.) s
5 313 M
( Changes from draft-irtf-tmrg-metrics-01a.txt:) s
5 291 M
( * Added to the discussion of packet drop rate metrics.) s
5 280 M
( Suggestions from Janardhan Iyengar, Sean Moore,) s
5 269 M
( Armando Caro, and Nelson Fonseca.) s
5 247 M
( * Added a sentence about throughput used as a metric for) s
5 236 M
( transfer times for very short flows.) s
5 225 M
( Response to email from Seam Moore.) s
5 203 M
( Changes from draft-irtf-tmrg-metrics-00.txt:) s
5 181 M
( * Added a list of relevant congestion control mechanisms to) s
5 170 M
( the abstract. Suggestion from Sean Moore.) s
5 126 M
(Floyd Section 1. [Page 4]) s
_R
S
%%Page: (5) 5
%%BeginPageSetup
_S
24 24 translate
/pagenum 5 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( * Added to the Introduction. Suggestion from Dado Colussi.) s
5 665 M
( * Added a sentence about jitter to the discussion of minimizing) s
5 654 M
( oscillations. Suggestion from Wesley Eddy.) s
5 632 M
( * Added a note about convergence between existing flows after) s
5 621 M
( a change in bandwidth. Suggestion from Wesley Eddy.) s
5 599 M
( * Added to the section on deployability. Suggestion from) s
5 588 M
( Wesley Eddy.) s
5 566 M
( Changes from draft-floyd-transport-metrics-00.txt:) s
5 544 M
( * Added metrics for:) s
5 533 M
( - robustness in challenging environments,) s
5 522 M
( - deployability,) s
5 511 M
( - robustness to failures and to misbehaving users) s
5 489 M
( * Added a discussion of fairness and packet size.) s
5 456 M
(2. Introduction) s
5 434 M
( As a step towards improving our methodologies for evaluating) s
5 423 M
( congestion control mechanisms, in this document we discuss some of) s
5 412 M
( the metrics to be considered. We also consider the relationship) s
5 401 M
( between metrics, e.g., the well-known tradeoff between throughput) s
5 390 M
( and delay.) s
5 368 M
( We consider metrics for aggregate traffic \(taking into account the) s
5 357 M
( effect of flows on competing traffic in the network\) as well as the) s
5 346 M
( heterogeneous goals of different applications or transport protocols) s
5 335 M
( \(e.g., of high throughput for bulk data transfer, and of low delay) s
5 324 M
( for interactive voice or video\). Different transport protocols or) s
5 313 M
( AQM mechanisms might have goals of optimizing different sets of) s
5 302 M
( metrics, with one transport protocol optimized for per-flow) s
5 291 M
( throughput and another optimized for robustness over wireless links,) s
5 280 M
( and with different degrees of attention to fairness with competing) s
5 269 M
( traffic. We hope this document will be used as a step in evaluating) s
5 258 M
( proposed congestion control mechanisms for a wide range of metrics,) s
5 247 M
( noting that Mechanism X is good at optimizing Metric A, but pays the) s
5 236 M
( price with poor performance for Metric B. The goal would be to have) s
5 225 M
( a broad view of both the strengths and weaknesses of newly-proposed) s
5 214 M
( congestion control mechanisms.) s
5 192 M
( Subsequent documents are planned to present sets of simulation and) s
5 181 M
( testbed scenarios for the evaluation of transport protocols and of) s
5 170 M
( congestion control mechanisms, based on the best current practice of) s
5 126 M
(Floyd Section 2. [Page 5]) s
_R
S
%%Page: (6) 6
%%BeginPageSetup
_S
24 24 translate
/pagenum 6 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( the research community. These are not intended to be complete or) s
5 676 M
( final benchmark test suites, but simply to be one step of many to be) s
5 665 M
( used by researchers in evaluating congestion control mechanisms.) s
5 654 M
( Subsequent documents are also planned on the methodologies in using) s
5 643 M
( these sets of scenarios.) s
5 621 M
( This is work from the Transport Modeling Research Group \(TMRG\) in) s
5 610 M
( the IRTF \(Internet Research Task Force\).) s
5 588 M
(3. Metrics) s
5 566 M
( The metrics that we discuss are the following:) s
5 544 M
( o Throughput;) s
5 522 M
( o Delay;) s
5 500 M
( o Packet loss rates;) s
5 478 M
( o Response to sudden changes or to transient events;) s
5 456 M
( o Minimizing oscillations in throughput or in delay;) s
5 434 M
( o Fairness and convergence times;) s
5 412 M
( o Robustness for challenging environments;) s
5 390 M
( o Robustness to failures and to misbehaving users;) s
5 368 M
( o Deployability;) s
5 346 M
( o Metrics for specific types of transport.) s
5 324 M
( o User-based metrics.) s
5 302 M
( We consider each of these below. Many of the metrics have network-) s
5 291 M
( based, flow-based, and user-based interpretations. For example,) s
5 280 M
( network-based metrics can consider aggregate bandwidth and aggregate) s
5 269 M
( drop rates, flow-based metrics can consider end-to-end transfer) s
5 258 M
( times for file transfers or end-to-end delay and packet drop rates) s
5 247 M
( for interactive traffic, and user-based metrics can consider user) s
5 236 M
( wait time or user satisfaction with the multimedia experience. Our) s
5 225 M
( main goal in this document is to explain the set of metrics that can) s
5 214 M
( be relevant, and not to legislate on the more appropriate) s
5 203 M
( methodology for using each general metric.) s
5 181 M
( For some of the metrics, such as fairness between flows, there is) s
5 170 M
( not a clear agreement in the network community about the desired) s
5 126 M
(Floyd Section 3. [Page 6]) s
_R
S
%%Page: (7) 7
%%BeginPageSetup
_S
24 24 translate
/pagenum 7 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( goals. In these cases, the document attempts to present the range) s
5 676 M
( of approaches.) s
5 654 M
(3.1. Throughput, Delay, and Loss Rates) s
5 632 M
( Because of the clear tradeoffs between throughput, delay, and loss) s
5 621 M
( rates, it can be useful to consider the three metrics together.) s
5 599 M
( An alternative would be to consider a separate metric such as power,) s
5 588 M
( defined in this context as throughput over delay, that combines) s
5 577 M
( throughput and delay. However, we do not propose in this document a) s
5 566 M
( clear target in terms of the tradeoffs between throughput and delay;) s
5 555 M
( we are simply proposing that the evaluation of transport protocols) s
5 544 M
( include an exploration of the competing metrics.) s
5 522 M
(3.1.1. Throughput) s
5 500 M
( Throughput can be measured as a router-based metric of aggregate) s
5 489 M
( link throughput, as a flow-based metric of per-connection transfer) s
5 478 M
( times, and as user-based metrics of utility functions or user wait) s
5 467 M
( times. It is a clear goal of most congestion control mechanisms to) s
5 456 M
( maximize throughput, subject to application demand and to the) s
5 445 M
( constraints of the other metrics.) s
5 423 M
( Throughput is sometimes distinguished from goodput, where throughput) s
5 412 M
( is the link or flow throughput in bytes per second, and goodput,) s
5 401 M
( also measured in bytes per second, is the subset of throughput) s
5 390 M
( consisting of useful traffic. That is, `goodput' excludes duplicate) s
5 379 M
( packets, packets that will be dropped downstream, packet fragments) s
5 368 M
( or ATM cells that are dropped at the receiver because they can't be) s
5 357 M
( re-assembled into complete packets, and the like.) s
5 335 M
( We note that maximizing throughput is of concern in a wide range of) s
5 324 M
( environments, from highly-congested networks to under-utilized ones,) s
5 313 M
( and from long-lived flows to very short ones. As an example,) s
5 302 M
( throughput has been used as one of the metrics for evaluating Quick-) s
5 291 M
( Start, a proposal to allow flows to start-up faster than slow-start,) s
5 280 M
( where throughput has been evaluated in terms of the transfer times) s
5 269 M
( for connections with a range of transfer sizes [QuickStart].) s
5 247 M
( In some contexts, it might be sufficient to consider the aggregate) s
5 236 M
( throughput or the mean per-flow throughput, while in other contexts) s
5 225 M
( it might be necessary to consider the distribution of per-flow) s
5 214 M
( throughput. Some researchers evaluate transport protocols in terms) s
5 203 M
( of maximizing the aggregate user utility, where a user's utility is) s
5 192 M
( generally defined as a function of the user's throughput [KMT98].) s
5 170 M
( Individual applications can have application-specific needs in terms) s
5 126 M
(Floyd Section 3.1.1. [Page 7]) s
_R
S
%%Page: (8) 8
%%BeginPageSetup
_S
24 24 translate
/pagenum 8 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( of throughput. For example, real-time video traffic can have highly) s
5 676 M
( variable bandwidth demands; VoIP traffic is sensitive to the amount) s
5 665 M
( of bandwidth received immediately after idle periods. Thus, user) s
5 654 M
( metrics for throughput can be more complex than simply the per-) s
5 643 M
( connection transfer time.) s
5 621 M
(3.1.2. Delay) s
5 599 M
( Like throughput, delay can be measured as a router-based metric of) s
5 588 M
( queueing delay over time, or as a flow-based metric in terms of per-) s
5 577 M
( packet transfer times. For reliable transfer, the per-packet) s
5 566 M
( transfer time includes the possible delay of retransmitting a lost) s
5 555 M
( packet.) s
5 533 M
( Users of bulk data transfer applications might care about per-packet) s
5 522 M
( transfer times only in so far as they affect the per-connection) s
5 511 M
( transfer time. On the other end of the spectrum, for users of) s
5 500 M
( streaming media, per-packet delay can be a significant concern.) s
5 489 M
( Note that in some cases the average delay might not capture the) s
5 478 M
( metric of interest to the users; for example, some users might care) s
5 467 M
( about the worst-case delay, or about the tail of the delay) s
5 456 M
( distribution.) s
5 434 M
(3.1.3. Packet Loss Rates) s
5 412 M
( Packet loss rates can be measured as a network-based or as a flow-) s
5 401 M
( based metric.) s
5 379 M
( When evaluating the effect of packet losses or ECN marks \(Explicit) s
5 368 M
( Congestion Notification, RFC 3168\) on the performance of a) s
5 357 M
( congestion control mechanism for an individual flow, researchers) s
5 346 M
( often use both the packet loss/mark rate for that connection, and) s
5 335 M
( the congestion event rate \(also called the loss event rate\), where a) s
5 324 M
( congestion event or loss event consists of one or more lost or) s
5 313 M
( marked packets in one round-trip time [RFC 3448].) s
5 291 M
( Some users might care about packet loss rates only in so far as they) s
5 280 M
( affect per-connection transfer times, while other users might care) s
5 269 M
( about packet loss rates directly. RFC 3611, RTP Control Protocol) s
5 258 M
( Extended Reports, describes a VoIP performance-reporting standard) s
5 247 M
( called RTCP XR, which includes a set of burst metrics. In RFC 3611,) s
5 236 M
( a burst is defined as the maximal sequence starting and ending with) s
5 225 M
( a lost packet, and not including a sequence of Gmin or more packets) s
5 214 M
( that are not lost [RFC 3611]. The burst metrics in RFC 3611 consist) s
5 203 M
( of the burst density \(the fraction of packets in bursts\), gap) s
5 192 M
( density \(the fraction of packets in the gaps between bursts\), burst) s
5 181 M
( duration \(the mean duration of bursts in seconds\), and gap duration) s
5 170 M
( \(the mean duration of gaps in seconds\).) s
5 126 M
(Floyd Section 3.1.3. [Page 8]) s
_R
S
%%Page: (9) 9
%%BeginPageSetup
_S
24 24 translate
/pagenum 9 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( In some cases it is useful to distinguish between packets dropped at) s
5 676 M
( routers due to congestion, and packets lost in the network due to) s
5 665 M
( corruption.) s
5 643 M
( One network-related reason to avoid high steady-state packet loss) s
5 632 M
( rates is to avoid congestion collapse in environments containing) s
5 621 M
( paths with multiple congested links. In such environments, high) s
5 610 M
( packet loss rates could result in congested links wasting scarce) s
5 599 M
( bandwidth by carrying packets that will only be dropped downstream,) s
5 588 M
( before being delivered to the receiver.) s
5 566 M
(3.2. Response Times and Minimizing Oscillations) s
5 544 M
( In this section we consider response times and oscillations) s
5 533 M
( together, as there are well-known tradeoffs between improving) s
5 522 M
( response times and minimizing oscillations. In addition, the) s
5 511 M
( scenarios that illustrate the dangers of poor response times are) s
5 500 M
( often quite different from the scenarios that illustrate the dangers) s
5 489 M
( of unnecessary oscillations.) s
5 467 M
(3.2.1. Response to Changes) s
5 445 M
( One of the key concerns in the design of congestion control) s
5 434 M
( mechanisms has been the response times to sudden congestion in the) s
5 423 M
( network. On the one hand, congestion control mechanisms should) s
5 412 M
( respond reasonably promptly to sudden congestion from routing or) s
5 401 M
( bandwidth changes, or from a burst of competing traffic. At the) s
5 390 M
( same time, congestion control mechanisms should not respond too) s
5 379 M
( severely to transient changes, e.g., to a sudden increase in delay) s
5 368 M
( that will dissipate in less than the connection's round-trip time.) s
5 346 M
( Evaluating the response to sudden or transient changes can be of) s
5 335 M
( particular concern for slowly-responding congestion control) s
5 324 M
( mechanisms such as equation-based congestion control [RFC 3448], and) s
5 313 M
( for AIMD \(Additive Increase Multiplicative Decrease\) or related) s
5 302 M
( mechanisms using parameters that make them more slowly-responding) s
5 291 M
( that TCP [BB01, BBFS01].) s
5 269 M
( In addition to the responsiveness and smoothness of aggregate) s
5 258 M
( traffic, one can consider the tradeoffs between responsiveness,) s
5 247 M
( smoothness, and aggressiveness for an individual connection [FHP00].) s
5 236 M
( In this case smoothness can be defined by the largest reduction in) s
5 225 M
( the sending rate in one round-trip time, in a deterministic) s
5 214 M
( environment with a packet drop exactly every 1/p packets. The) s
5 203 M
( responsiveness is defined as the number of round-trip times of) s
5 192 M
( sustained congested required for the sender to halve the sending) s
5 181 M
( rate, and the aggressiveness is defined as the maximum increase in) s
5 170 M
( the sending rate in one round-trip time, in packets per second, in) s
5 126 M
(Floyd Section 3.2.1. [Page 9]) s
_R
S
%%Page: (10) 10
%%BeginPageSetup
_S
24 24 translate
/pagenum 10 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( the absence of congestion.) s
5 665 M
(3.2.2. Minimizing Oscillations) s
5 643 M
( One goal is that of stability, in terms of minimizing oscillations) s
5 632 M
( of queueing delay or of throughput. Scenarios illustrating) s
5 621 M
( oscillations are often dominated by long-lived connections, perhaps) s
5 610 M
( with a small number of changes in the level of congestion.) s
5 588 M
( Minimizing oscillations in queueing delay or throughput has related) s
5 577 M
( per-flow metrics of minimizing jitter in round-trip times and loss) s
5 566 M
( rates.) s
5 544 M
( An orthogonal goal for some congestion control mechanisms, e.g., for) s
5 533 M
( equation-based congestion control, is to minimize the oscillations) s
5 522 M
( in the sending rate for an individual connection, given an) s
5 511 M
( environment with a fixed, steady-state packet drop rate. \(As is) s
5 500 M
( well known, TCP congestion control is characterized by a pronounced) s
5 489 M
( oscillation in the sending rate, with the sender halving the sending) s
5 478 M
( rate in response to congestion.\) One metric for the level of) s
5 467 M
( oscillations is the smoothness metric given above.) s
5 445 M
(3.3. Fairness and Convergence) s
5 423 M
( Another set of metrics are those of fairness and of convergence) s
5 412 M
( times. Fairness can be considered between flows of the same) s
5 401 M
( protocol, and between flows using different protocols \(e.g.,) s
5 390 M
( fairness between TCP and a new transport protocol\).) s
5 368 M
( There are a number of different fairness measures. These include) s
5 357 M
( max-min fairness [HG86], proportional fairness [KMT98, K01], the) s
5 346 M
( fairness index proposed in [JCH84], and the product measure, a) s
5 335 M
( variant of network power [BJ81].) s
5 313 M
( Max-min fairness: In order to satisfy the max-min fairness criteria,) s
5 302 M
( the smallest throughput rate must be as large as possible. Given) s
5 291 M
( this condition, the next-smallest throughput rate must be as large) s
5 280 M
( as possible, and so on. Thus, the max-min fairness gives absolute) s
5 269 M
( priority to the smallest flows.) s
5 247 M
( Epsilon-fairness: A metric related to max-min fairness is epsilon-) s
5 236 M
( fairness, where a rate allocation is defined as epsilon-fair if) s
5 214 M
( min_i x_i / max_i x_i >= 1 - epsilon.) s
5 192 M
( where x_i is the resource allocation to the i-th user. Epsilon-) s
5 181 M
( fairness measures the worst-case ratio between any two throughput) s
5 170 M
( rates [ZKL04].) s
5 126 M
(Floyd Section 3.3. [Page 10]) s
_R
S
%%Page: (11) 11
%%BeginPageSetup
_S
24 24 translate
/pagenum 11 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( Proportional fairness: In contrast, an allocation x is defined as) s
5 676 M
( proportionally fair if for any other feasible allocation x*, the) s
5 665 M
( aggregate of proportional changes is zero or negative:) s
5 643 M
( sum_i \(x*_i - x_i\)/x_i <= 0.) s
5 621 M
( "This criterion favours smaller flows, but less emphatically than) s
5 610 M
( max-min fairness" [K01].) s
5 588 M
( Jain's fairness index: The fairness index in [JCH84] is) s
5 566 M
( \(\( sum_i x_i \)^2\) / \(n * sum_i \(x_i\)^2 \) ,) s
5 544 M
( where there are n users. This fairness index ranges from 0 to 1,) s
5 533 M
( and is maximum when all users receive the same allocation. This) s
5 522 M
( index is k/n when k users equally share the resource, and the other) s
5 511 M
( n-k users receive zero allocation.) s
5 489 M
( The product measure: The product measure) s
5 467 M
( product_i x_i ,) s
5 445 M
( the product of the throughput of the individual connections, is also) s
5 434 M
( used as a measure of fairness. For our purposes, let x_i be the) s
5 423 M
( throughput for the i-th connection. \(In other contexts x_i is taken) s
5 412 M
( as the power of the i-th connection, and the product measure is) s
5 401 M
( referred to as network power.\) The product measure is particularly) s
5 390 M
( sensitive to segregation; the product measure is zero if any) s
5 379 M
( connection receives zero throughput. In [MS91] it is shown that for) s
5 368 M
( a network with many connections and one shared gateway, the product) s
5 357 M
( measure is maximized when all connections receive the same) s
5 346 M
( throughput.) s
5 324 M
( In [CRM05], Colussi et al. propose a new definition of fairness,) s
5 313 M
( that "a set of TCP fair flows do not cause more congestion than a) s
5 302 M
( set of TCP flows would cause", where congestion is defined in terms) s
5 291 M
( of queueing delay, queueing delay variation, the congestion event) s
5 280 M
( rate [e.g., loss event rate], and the packet loss rate.) s
5 258 M
( Chiu and Tan in [CT06] argue for redefining the notion of fairness) s
5 247 M
( when studying traffic controls for inelastic traffic, proposing that) s
5 236 M
( inelastic flows adopt other traffic controls such as admission) s
5 225 M
( control.) s
5 203 M
( Fairness and the number of congested links: Some of these fairness) s
5 192 M
( metrics are discussed in more detail in [F91]. We note that there) s
5 181 M
( is not a clear consensus for the fairness goals, in particular for) s
5 170 M
( fairness between flows that traverse different numbers of congested) s
5 126 M
(Floyd Section 3.3. [Page 11]) s
_R
S
%%Page: (12) 12
%%BeginPageSetup
_S
24 24 translate
/pagenum 12 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( links [F91].) s
5 665 M
( Fairness and round-trip times: One goal cited in a number of new) s
5 654 M
( transport protocols has been that of fairness between flows with) s
5 643 M
( different round-trip times [KHR02, XHR04]. We note that there is not) s
5 632 M
( a consensus in the networking community about the desirability of) s
5 621 M
( this goal, or about the implications and interactions between this) s
5 610 M
( goal and other metrics [FJ92] \(Section 3.3\).) s
5 588 M
( Fairness and packet size: One fairness issue is that of the relative) s
5 577 M
( fairness for flows with different packet sizes. Many file transfer) s
5 566 M
( applications will use the maximum packet size possible; in) s
5 555 M
( contrast, low-bandwidth VoIP flows are likely to send small packets,) s
5 544 M
( sending a new packet every 10 to 40 ms., to limit delay. Should a) s
5 533 M
( small-packet VoIP connection receive the same sending rate in bytes) s
5 522 M
( per second as a large-packet TCP connection in the same environment,) s
5 511 M
( or should it receive the same sending rate in *packets* per second?) s
5 500 M
( This fairness issue has been discussed in more detail in [FK04],) s
5 489 M
( with [FK05] also describing the ways that packet size can effect the) s
5 478 M
( packet drop rate experienced by a flow.) s
5 456 M
( Convergence times: Convergence times concern the time for) s
5 445 M
( convergence to fairness between an existing flow and a newly-) s
5 434 M
( starting one, and are a special concern for environments with high-) s
5 423 M
( bandwidth flows. Convergence times also concern the time for) s
5 412 M
( convergence to fairness between two existing flows after a sudden) s
5 401 M
( change such as a change in link capacity on a wireless link. As) s
5 390 M
( with fairness, convergence times can matter both between flows of) s
5 379 M
( the same protocol, and between flows using different protocols) s
5 368 M
( [SLFK03]. One metric used for convergence times is the delta-fair) s
5 357 M
( convergence time, defined as the time taken for two flows with the) s
5 346 M
( same round-trip time to go from shares of 100/101-th and 1/101-th of) s
5 335 M
( the link bandwidth, to having close to fair sharing with shares of) s
5 324 M
( \(1+delta\)/2 and \(1-delta\)/2 of the link bandwidth [BBFS01]. A) s
5 313 M
( similar metric for convergence times measures the convergence time) s
5 302 M
( as the number of round-trip times for two flows to reach epsilon-) s
5 291 M
( fairness, when starting from a maximally-unfair state [ZKL04]. ') s
5 258 M
(3.4. Robustness for Challenging Environments) s
5 236 M
( While congestion control mechanisms are generally evaluated first) s
5 225 M
( over environments with static routing in a network of two-way point-) s
5 214 M
( to-point links, some environments bring up more challenging problems) s
5 203 M
( \(e.g., corrupted packets, variable bandwidth, mobility\) as well as) s
5 192 M
( new metrics to be considered \(e.g., energy consumption\).) s
5 170 M
( Robustness for challenging environments: Robustness needs to be) s
5 126 M
(Floyd Section 3.4. [Page 12]) s
_R
S
%%Page: (13) 13
%%BeginPageSetup
_S
24 24 translate
/pagenum 13 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( explored for paths with reordering, corruption, variable bandwidth,) s
5 676 M
( asymmetric routing, router configuration changes, mobility, and the) s
5 665 M
( like. In general, Internet architecture has valued robustness over) s
5 654 M
( efficiency, e.g., when there are tradeoffs between robustness and) s
5 643 M
( the throughput, delay, and fairness metrics described above.) s
5 621 M
( Energy consumption: In mobile environments the energy consumption) s
5 610 M
( for the mobile end-node can be a key metric that is affected by the) s
5 599 M
( transport protocol [TM02].) s
5 577 M
( Goodput: For wireless networks, goodput can be a useful metric,) s
5 566 M
( where goodput is defined as the fraction of useful data from all of) s
5 555 M
( the data delivered. High goodput indicates an efficient use of the) s
5 544 M
( radio spectrum and lower interference to other users [GF04].) s
5 522 M
(3.5. Robustness to Failures and to Misbehaving Users) s
5 500 M
( One goal is for congestion control mechanisms to be robust to) s
5 489 M
( misbehaving users, such as receivers that `lie' to data senders) s
5 478 M
( about the congestion experienced along the path or otherwise attempt) s
5 467 M
( to bypass the congestion control mechanisms of the sender [SCWA99].) s
5 456 M
( Another goal is for congestion control mechanisms to be as robust as) s
5 445 M
( possible to failures, such as failures of routers in using explicit) s
5 434 M
( feedback to end-nodes or failures of end-nodes to follow the) s
5 423 M
( prescribed protocols,) s
5 401 M
(3.6. Deployability) s
5 379 M
( One metric for congestion control mechanisms is their deployability) s
5 368 M
( in the current Internet. Metrics related to deployability include) s
5 357 M
( the ease of failure diagnosis, and the overhead in terms of packet) s
5 346 M
( header size or added complexity at end-nodes or routers.) s
5 324 M
( One key aspect of deployability concerns the range of deployment) s
5 313 M
( needed for a new congestion control mechanism. Consider the) s
5 302 M
( following possible deployment requirements:) s
5 280 M
( * Only at the sender \(e.g., NewReno in TCP\);) s
5 269 M
( * Only at the receiver \(e.g., delayed acknowledgements in TCP\);) s
5 258 M
( * Both the sender and receiver \(e.g., SACK TCP\);) s
5 247 M
( * At a single router \(e.g., RED\);) s
5 236 M
( * All of the routers along the end-to-end path;) s
5 225 M
( * Both end nodes and all routers along the path \(e.g., XCP\).) s
5 203 M
( Another deployability issue concerns the complexity of the code.) s
5 192 M
( Roughly how many lines of code are required to implement the) s
5 181 M
( mechanism in software? Is floating point math required? We note) s
5 170 M
( that we don't suggest these questions as ways to reduce the) s
5 126 M
(Floyd Section 3.6. [Page 13]) s
_R
S
%%Page: (14) 14
%%BeginPageSetup
_S
24 24 translate
/pagenum 14 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( deployability metric to a single number; we suggest them as issues) s
5 676 M
( that could be considered in evaluating the deployability of a) s
5 665 M
( proposed congestion control mechanism.) s
5 643 M
(3.7. Metrics for Specific Types of Transport) s
5 621 M
( In some cases modified metrics are needed for evaluting transport) s
5 610 M
( protocols intended for QoS-enabled environments or for below-best-) s
5 599 M
( effort traffic [VKD02, KK03].) s
5 577 M
(3.8. User-Based Metrics) s
5 555 M
( An alternate approach that has been proposed for the evaluation of) s
5 544 M
( congestion control mechanisms would be to evaluate in terms of user) s
5 533 M
( metrics such as user satisfaction, or in terms of application-) s
5 522 M
( specific utility functions. Such an approach would require the) s
5 511 M
( definition of a range of user metrics or of application-specific) s
5 500 M
( utility functions for the range of applications under consideration) s
5 489 M
( \(e.g., FTP, HTTP, VoIP\).) s
5 467 M
(4. Metrics in the IP Performance Metrics \(IPPM\) Working Group) s
5 445 M
( The IPPM Working Group [IPPM] was established to define performance) s
5 434 M
( metrics to be used by network operators, end users, or independent) s
5 423 M
( testing groups. The metrics include metrics for connectivity, delay) s
5 412 M
( and loss, delay variation, loss patterns, packet reordering, bulk) s
5 401 M
( transfer capacity, and link capacity. The IPPM documents give) s
5 390 M
( concrete, well-defined metrics, along with a methodology for) s
5 379 M
( measuring the metric. The metrics discussed in this document have a) s
5 368 M
( different purpose from the IPPM metrics; in this document we are) s
5 357 M
( discussing metrics as used in analysis, simulations, and experiments) s
5 346 M
( for the evaluation of congestion control mechanisms. Further, we) s
5 335 M
( are discussing these metrics in a general sense, rather than looking) s
5 324 M
( for specific concrete definitions for each metric. However, there) s
5 313 M
( are many cases where the metric definitions from IPPM could be) s
5 302 M
( useful, particularly for specific issues of how to measure these) s
5 291 M
( metrics in simulations or in testbeds.) s
5 269 M
(5. Comments on Methodology) s
5 247 M
( The types of scenarios that are used to test specific metrics, and) s
5 236 M
( the range of parameters that it is useful to consider, will be) s
5 225 M
( discussed in separate documents, e.g., along with specific scenarios) s
5 214 M
( for use in evaluating congestion control mechanisms.) s
5 192 M
( We note that it can be important to evaluate metrics over a wide) s
5 181 M
( range of environments, with a range of link bandwidths, congestion) s
5 170 M
( levels, and levels of statistical multiplexing. It is also) s
5 126 M
(Floyd Section 5. [Page 14]) s
_R
S
%%Page: (15) 15
%%BeginPageSetup
_S
24 24 translate
/pagenum 15 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( important to evaluate congestion control mechanisms in a range of) s
5 676 M
( scenarios, including typical ranges of connection sizes and round-) s
5 665 M
( trip times [FK02]. It is also useful to compare metrics for new or) s
5 654 M
( modified transport protocols with those of the current standards for) s
5 643 M
( TCP.) s
5 621 M
( Li et al. in "Experimental Evaluation of TCP Protocols for High-) s
5 610 M
( Speed Networks" [LLS05] focus on the performance of TCP in high-) s
5 599 M
( speed networks, and consider metrics for aggregate throughput, loss) s
5 588 M
( rates, fairness \(including fairness between flows with different) s
5 577 M
( round-trip times\), response times \(including convergence times\), and) s
5 566 M
( incremental deployment.) s
5 544 M
( More general references on methodology include [J91]. Papers that) s
5 533 M
( discuss the range of metrics for evaluating congestion control) s
5 522 M
( include [MTZ04].) s
5 500 M
(6. Security Considerations) s
5 478 M
( There are no security considerations in this document.) s
5 456 M
(7. IANA Considerations) s
5 434 M
( There are no IANA considerations in this document.) s
5 412 M
(8. Acknowledgements) s
5 390 M
( Thanks to Armando Caro, Dah Ming Chiu, Dado Colussi, Wesley Eddy,) s
5 379 M
( Nelson Fonseca, Janardhan Iyengar, Doug Leith, Saverio Mascolo, Sean) s
5 368 M
( Moore, and Damon Wischik, and members of the Transport Modeling) s
5 357 M
( Research Group for feedback and contributions.) s
5 335 M
(Informative References) s
5 313 M
( [BB01] D. Bansal and H. Balakrishnan, Binomial Congestion Control) s
5 302 M
( Algorithms, IEEE Infocom, April 2001.) s
5 280 M
( [BBFS01] D. Bansal, H. Balakrishnan, S. Floyd, and S. Shenker,) s
5 269 M
( Dynamic Behavior of Slowly-Responsive Congestion Control) s
5 258 M
( Algorithms, SIGCOMM 2001.) s
5 236 M
( [BJ81] K. Bharath-Kumar and J. Jeffrey, A New Approach to) s
5 225 M
( Performance-Oriented Flow Control, IEEE Transactions on) s
5 214 M
( Communications, Vol.COM-29 N.4, April 1981.) s
5 192 M
( [CRM05] A New Approach to TCP-Fairness, Report C-2005-49, University) s
5 181 M
( of Helsinki, Finland, 2005.) s
5 126 M
(Floyd [Page 15]) s
_R
S
%%Page: (16) 16
%%BeginPageSetup
_S
24 24 translate
/pagenum 16 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( [CT06] D. Chiu and A. Tam, Redefining Fairness in the Study of TCP-) s
5 676 M
( friendly Traffic Controls, Technical Report, 2006.) s
5 654 M
( [F91] S. Floyd, Connections with Multiple Congested Gateways in) s
5 643 M
( Packet-Switched Networks Part 1: One-way Traffic, Computer) s
5 632 M
( Communication Review, Vol.21, No.5, October 1991, p. 30-47.) s
5 610 M
( [FK05] S. Floyd and E. Kohler, TFRC for Voice: the VoIP Variant,) s
5 599 M
( draft-ietf-dccp-tfrc-voip-02.txt, internet draft, work in) s
5 588 M
( progress, July 2005.) s
5 566 M
( [FHP00] S. Floyd, M. Handley, and J. Padhye, A Comparison of) s
5 555 M
( Equation-Based and AIMD Congestion Control, May 2000. URL) s
5 544 M
( "http://www.icir.org/tfrc/".) s
5 522 M
( [FJ92] S. Floyd and V. Jacobson, On Traffic Phase Effects in Packet-) s
5 511 M
( Switched Gateways, Internetworking: Research and Experience, V.3) s
5 500 M
( N.3, September 1992, p.115-156.) s
5 478 M
( [FK04] S. Floyd and J. Kempf, IAB Concerns Regarding Congestion) s
5 467 M
( Control for Voice Traffic in the Internet, RFC 3714, March 2004.) s
5 445 M
( [FK02] S. Floyd and E. Kohler, Internet Research Needs Better) s
5 434 M
( Models, Hotnets-I. October 2002.) s
5 412 M
( [GF04] A. Gurtov and S. Floyd, Modeling Wireless Links for Transport) s
5 401 M
( Protocols, ACM CCR, 34\(2\):85-96, April 2004.) s
5 379 M
( [HG86] E. Hahne and R. Gallager, Round Robin Scheduling for Fair) s
5 368 M
( Flow Control in Data Communications Networks, IEEE International) s
5 357 M
( Conference on Communications, June 1986.) s
5 335 M
( [IPPM] IP Performance Metrics \(IPPM\) Working Group, URL) s
5 324 M
( "http://www.ietf.org/html.charters/ippm-charter.html".) s
5 302 M
( [J91] R. Jain, The Art of Computer Systems Performance Analysis:) s
5 291 M
( Techniques for Experimental Design, Measurement, Simulation, and) s
5 280 M
( Modeling, John Wiley & Sons, 1991.) s
5 258 M
( [JCH84] R. Jain, D.M. Chiu, and W. Hawe, A Quantitative Measure of) s
5 247 M
( Fairness and Discrimination for Resource Allocation in Shared) s
5 236 M
( Systems, DEC TR-301, Littleton, MA: Digital Equipment) s
5 225 M
( Corporation, 1984.) s
5 203 M
( [K01] F. Kelly, Mathematical Modelling of the Internet, "Mathematics) s
5 192 M
( Unlimited - 2001 and Beyond" \(Editors B. Engquist and W.) s
5 181 M
( Schmid\), Springer-Verlag, Berlin, pp. 685-702, 2001.) s
5 126 M
(Floyd [Page 16]) s
_R
S
%%Page: (17) 17
%%BeginPageSetup
_S
24 24 translate
/pagenum 17 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( [KHR02] D. Katabi, M. Handley, and C. Rohrs, Congestion Control for) s
5 676 M
( High Bandwidth-Delay Product Networks, ACM Sigcomm, 2002.) s
5 654 M
( [KK03] A. Kuzmanovic and E. W. Knightly, TCP-LP: A Distributed) s
5 643 M
( Algorithm for Low Priority Data Transfer, IEEE INFOCOM 2003,) s
5 632 M
( April 2003.) s
5 610 M
( [KMT98] F. Kelly, A. Maulloo and D. Tan, Rate Control in) s
5 599 M
( Communication Networks: Shadow Prices, Proportional Fairness and) s
5 588 M
( Stability. Journal of the Operational Research Society 49, pp.) s
5 577 M
( 237-252, 1998.) s
5 555 M
( [LLS05] Y-T. Li, D. Leith, and R. Shorten, Experimental Evaluation) s
5 544 M
( of TCP Protocols for High-Speed Networks, Hamilton Institute,) s
5 533 M
( 2005. URL "http://www.hamilton.ie/net/eval/results-HI2005.pdf".) s
5 511 M
( [MS91] D. Mitra and J. Seery, Dynamic Adaptive Windows for High) s
5 500 M
( Speed Data Networks with Multiple Paths and Propagation Delays,) s
5 489 M
( INFOCOM '91, pp 39-48. [MTZ04] L. Mamatas, V. Tsaoussidis, and) s
5 478 M
( C. Zhang, Approaches to Congestion Control in Packet Networks,) s
5 467 M
( 2004.) s
5 445 M
( [QuickStart] Quick-Start Web Page, URL) s
5 434 M
( "http://www.icir.org/floyd/quickstart.html".) s
5 412 M
( [RFC 2119] S. Bradner. Key Words For Use in RFCs to Indicate) s
5 401 M
( Requirement Levels. RFC 2119.) s
5 379 M
( BIBREF RFC3168 "RFC 3168" K Ramakrishnan, S. Floyd, and D. Black,) s
5 368 M
( The Addition of Explicit Congestion Notification \(ECN\) to IP,) s
5 357 M
( RFC 3168, September 2001.) s
5 335 M
( [RFC 3448] M. Handley, S. Floyd, J. Padhye, and J. Widmer, TCP) s
5 324 M
( Friendly Rate Control \(TFRC\): Protocol Specification, RFC 3448,) s
5 313 M
( Proposed Standard, January 2003.) s
5 291 M
( [RFC 3611] T. Friedman, R. Caceres, and A. Clark, RTP Control) s
5 280 M
( Protocol Extended Reports \(RTCP XR\), RFC 3611, November 2003.) s
5 258 M
( [SLFK03] R.N. Shorten, D.J. Leith, J. Foy, and R. Kilduff, Analysis) s
5 247 M
( and Design of Congestion Control in Synchronised Communication) s
5 236 M
( Networks. Proc. 12th Yale Workshop on Adaptive & Learning) s
5 225 M
( Systems, May 2003.) s
5 203 M
( [SCWA99] TCP Congestion Control with a Misbehaving Receiver, ACM) s
5 192 M
( Computer Communications Review, October 1999.) s
5 126 M
(Floyd [Page 17]) s
_R
S
%%Page: (18) 18
%%BeginPageSetup
_S
24 24 translate
/pagenum 18 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( [TM02] V. Tsaoussidis and I. Matta, Open Issues of TCP for Mobile) s
5 676 M
( Computing, Journal of Wireless Communications and Mobile) s
5 665 M
( Computing: Special Issue on Reliable Transport Protocols for) s
5 654 M
( Mobile Computing, February 2002.) s
5 632 M
( [VKD02] A. Venkataramani, R. Kokku, and M. Dahlin, TCP Nice: A) s
5 621 M
( Mechanism for Background Transfers, Fifth USENIX Symposium on) s
5 610 M
( Operating System Design and Implementation \(OSDI\), 2002.) s
5 588 M
( [XHR04] L. Xu, K. Harfoush, and I. Rhee, Binary Increase Congestion) s
5 577 M
( Control for Fast, Long Distance Networks, Infocom 2004.) s
5 555 M
( [ZKL04] Y. Zhang, S.-R. Kang, and D. Loguinov, Delayed Stability and) s
5 544 M
( Performance of Distributed Congestion Control, ACM SIGCOMM,) s
5 533 M
( August 2004.) s
5 511 M
(Authors' Addresses) s
5 489 M
( Sally Floyd <floyd@icir.org>) s
5 478 M
( ICSI Center for Internet Research) s
5 467 M
( 1947 Center Street, Suite 600) s
5 456 M
( Berkeley, CA 94704) s
5 445 M
( USA) s
5 412 M
(Full Copyright Statement) s
5 390 M
( Copyright \(C\) The Internet Society 2006. This document is subject) s
5 379 M
( to the rights, licenses and restrictions contained in BCP 78, and) s
5 368 M
( except as set forth therein, the authors retain all their rights.) s
5 346 M
( This document and the information contained herein are provided on) s
5 335 M
( an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE) s
5 324 M
( REPRESENTS OR IS SPONSORED BY \(IF ANY\), THE INTERNET SOCIETY AND THE) s
5 313 M
( INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR) s
5 302 M
( IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF) s
5 291 M
( THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED) s
5 280 M
( WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.) s
5 258 M
(Intellectual Property) s
5 236 M
( The IETF takes no position regarding the validity or scope of any) s
5 225 M
( Intellectual Property Rights or other rights that might be claimed) s
5 214 M
( to pertain to the implementation or use of the technology described) s
5 203 M
( in this document or the extent to which any license under such) s
5 192 M
( rights might or might not be available; nor does it represent that) s
5 181 M
( it has made any independent effort to identify any such rights.) s
5 170 M
( Information on the procedures with respect to rights in RFC) s
5 126 M
(Floyd [Page 18]) s
_R
S
%%Page: (19) 19
%%BeginPageSetup
_S
24 24 translate
/pagenum 19 def
/fname (metrics.txt) def
/fdir () def
/ftail (metrics.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: December 2006 June 2006) s
5 687 M
( documents can be found in BCP 78 and BCP 79.) s
5 665 M
( Copies of IPR disclosures made to the IETF Secretariat and any) s
5 654 M
( assurances of licenses to be made available, or the result of an) s
5 643 M
( attempt made to obtain a general license or permission for the use) s
5 632 M
( of such proprietary rights by implementers or users of this) s
5 621 M
( specification can be obtained from the IETF on-line IPR repository) s
5 610 M
( at http://www.ietf.org/ipr.) s
5 588 M
( The IETF invites any interested party to bring to its attention any) s
5 577 M
( copyrights, patents or patent applications, or other proprietary) s
5 566 M
( rights that may cover technology that may be required to implement) s
5 555 M
( this standard. Please address the information to the IETF at ietf-) s
5 544 M
( ipr@ietf.org.) s
5 126 M
(Floyd [Page 19]) s
_R
S
%%Trailer
%%Pages: 19
%%DocumentNeededResources: font Courier-Bold Courier
%%EOF
| PAFTECH AB 2003-2026 | 2026-04-23 11:05:18 |