One document matched: draft-irtf-tmrg-tools-01.ps
%!PS-Adobe-3.0
%%BoundingBox: 24 24 588 768
%%Title: Enscript Output
%%For: Sally Floyd
%%Creator: GNU enscript 1.6.1
%%CreationDate: Fri Oct 14 13:10:22 2005
%%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-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 731 M
(Internet Engineering Task Force S. Floyd) s
5 720 M
(INTERNET-DRAFT E. Kohler) s
5 709 M
(draft-irtf-tmrg-tools-01.txt Editors) s
5 698 M
(Expires: April 2006 14 October 2005) s
5 665 M
( Tools for the Evaluation of Simulation and Testbed Scenarios) 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 April 2006.) s
5 335 M
(Copyright Notice) s
5 313 M
( Copyright \(C\) The Internet Society \(2005\). All Rights Reserved.) s
5 291 M
(Abstract) s
5 269 M
( This document describes tools for the evaluation of simulation and) s
5 258 M
( testbed scenarios used in research on Internet congestion control) s
5 247 M
( mechanisms. We believe that research in congestion control) s
5 203 M
(Floyd, Kohler [Page 1]) s
_R
S
%%Page: (2) 2
%%BeginPageSetup
_S
24 24 translate
/pagenum 2 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( mechanisms has been seriously hampered by the lack of good models) s
5 676 M
( underpinning analysis, simulation, and testbed experiments, and that) s
5 665 M
( tools for the evaluation of simulation and testbed scenarios can) s
5 654 M
( help in the construction of better scenarios, based on better) s
5 643 M
( underlying models. One use of the tools described in this document) s
5 632 M
( is in comparing key characteristics of test scenarios with known) s
5 621 M
( characteristics from the diverse and ever-changing real world.) s
5 610 M
( Tools characterizing the aggregate traffic on a link include the) s
5 599 M
( distribution of per-packet round-trip times, the distribution of) s
5 588 M
( packet sequence numbers, and the like. Tools characterizing end-to-) s
5 577 M
( end paths include drop rates as a function of packet size and of) s
5 566 M
( burst size, the synchronization ratio between two end-to-end TCP) s
5 555 M
( flows, and the like. For each characteristic, we describe what) s
5 544 M
( aspects of the scenario determine this characteristic, how the) s
5 533 M
( characteristic can affect the results of simulations and experiments) s
5 522 M
( for the evaluation of congestion control mechanisms, and what is) s
5 511 M
( known about this characteristic in the real world. We also explain) s
5 500 M
( why the use of such tools can add considerable power to our) s
5 489 M
( understanding and evaluation of simulation and testbed scenarios.) s
5 126 M
(Floyd, Kohler [Page 2]) s
_R
S
%%Page: (3) 3
%%BeginPageSetup
_S
24 24 translate
/pagenum 3 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( Table of Contents) s
5 665 M
( 1. Introduction. . . . . . . . . . . . . . . . . . . . . . . 3) s
5 654 M
( 2. Conventions . . . . . . . . . . . . . . . . . . . . . . . 4) s
5 643 M
( 3. Tools . . . . . . . . . . . . . . . . . . . . . . . . . . 4) s
5 632 M
( 3.1. Characterizing Aggregate Traffic on a Link . . . . . 4) s
5 621 M
( 3.2. Characterizing an End-to-End Path. . . . . . . . . . 5) s
5 610 M
( 3.3. Other Characteristics. . . . . . . . . . . . . . . . 5) s
5 599 M
( 4. Distribution of per-packet round-trip times . . . . . . . 5) s
5 588 M
( 5. Distribution of packet sequence numbers . . . . . . . . . 7) s
5 577 M
( 6. The Distribution of Packet Sizes. . . . . . . . . . . . . 8) s
5 566 M
( 7. The Ratio Between Forward-path and Reverse-path Traf-) s
5 555 M
( fic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8) s
5 544 M
( 8. The Distribution of Per-Packet Peak Flow Rates. . . . . . 9) s
5 533 M
( 9. The Distribution of Transport Protocols.. . . . . . . . . 10) s
5 522 M
( 10. The Synchronization Ratio. . . . . . . . . . . . . . . . 10) s
5 511 M
( 11. Drop or Mark Rates as a Function of Packet Size. . . . . 12) s
5 500 M
( 12. Drop Rates as a Function of Burst Size.. . . . . . . . . 13) s
5 489 M
( 13. Drop Rates as a Function of Sending Rate.. . . . . . . . 15) s
5 478 M
( 14. Congestion Control Mechanisms for Traffic, along) s
5 467 M
( with Sender and. . . . . . . . . . . . . . . . . . . . . . . 16) s
5 456 M
( 15. Characterization of Congested Links in Terms of) s
5 445 M
( Bandwidth and Typical Levels of Congestion . . . . . . . . . 16) s
5 434 M
( 16. Characterization of Challenging Lower Layers.. . . . . . 16) s
5 423 M
( 17. Characterization of Network Changes Affecting Con-) s
5 412 M
( gestion. . . . . . . . . . . . . . . . . . . . . . . . . . . 16) s
5 401 M
( 18. Using the Tools Presented in this Document . . . . . . . 16) s
5 390 M
( 19. Related Work . . . . . . . . . . . . . . . . . . . . . . 16) s
5 379 M
( 20. Conclusions. . . . . . . . . . . . . . . . . . . . . . . 16) s
5 368 M
( 21. Security Considerations. . . . . . . . . . . . . . . . . 17) s
5 357 M
( 22. IANA Considerations. . . . . . . . . . . . . . . . . . . 17) s
5 346 M
( 23. Acknowledgements . . . . . . . . . . . . . . . . . . . . 17) s
5 335 M
( Informative References . . . . . . . . . . . . . . . . . . . 17) s
5 324 M
( Editors' Addresses . . . . . . . . . . . . . . . . . . . . . 19) s
5 313 M
( Full Copyright Statement . . . . . . . . . . . . . . . . . . 19) s
5 302 M
( Intellectual Property. . . . . . . . . . . . . . . . . . . . 19) s
5 280 M
(1. Introduction) s
5 258 M
( This document discusses tools for the evaluation of simulation and) s
5 247 M
( testbed scenarios used in research on Internet congestion control) s
5 236 M
( mechanisms. These tools include but are not limited to measurement) s
5 225 M
( tools; the tools discussed in this document are largely ways of) s
5 214 M
( characterizing aggregate traffic on a link, or characterizing the) s
5 203 M
( end-to-end path. One use of these tools is for understanding key) s
5 192 M
( characteristics of test scenarios; many characteristics, such as the) s
5 181 M
( distribution of per-packet round-trip times on the link, don't come) s
5 170 M
( from a single input parameter but are determined by a range of) s
5 126 M
(Floyd, Kohler Section 1. [Page 3]) s
_R
S
%%Page: (4) 4
%%BeginPageSetup
_S
24 24 translate
/pagenum 4 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( inputs. A second use of the tools is to compare key characteristics) s
5 676 M
( of test scenarios with what is known of the same characteristics of) s
5 665 M
( the past and current Internet, and with what can be conjectured) s
5 654 M
( about these characteristics of future networks. This paper follows) s
5 643 M
( the general approach from "Internet Research Needs Better Models") s
5 632 M
( [FK02].) s
5 610 M
( As an example of the power of tools for characterizing scenarios, a) s
5 599 M
( great deal is known about the distribution of connection sizes on a) s
5 588 M
( link, or equivalently, the distribution of per-packet sequence) s
5 577 M
( numbers. It has been conjectured that a heavy-tailed distribution) s
5 566 M
( of connection sizes is an invariant feature of Internet traffic. A) s
5 555 M
( test scenario with mostly long-lived traffic, or with a mix with) s
5 544 M
( only long-lived and very short flows, does not have a realistic) s
5 533 M
( distribution of connection sizes, and can give unrealistic results) s
5 522 M
( in simulations or experiments evaluating congestion control) s
5 511 M
( mechanisms. For instance, the distribution of packet sequence) s
5 500 M
( numbers makes clear the fraction of traffic on a link from medium-) s
5 489 M
( sized connections, e.g., with packet sequence numbers from 100 to) s
5 478 M
( 1000. These medium-sized connections can slow-start up to a large) s
5 467 M
( congestion window, possibly coming to an abrupt stop soon) s
5 456 M
( afterwards, contributing significantly to the burstiness of the) s
5 445 M
( aggregate traffic, and to the problems facing congestion control.) s
5 423 M
( In the sections below we will discuss a number of tools for) s
5 412 M
( describing and evaluating scenarios, show how these characteristics) s
5 401 M
( can affect the results of research on congestion control mechanisms,) s
5 390 M
( and summarize what is known about these characteristics in real-) s
5 379 M
( world networks.) s
5 357 M
(2. Conventions) s
5 335 M
( The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",) s
5 324 M
( "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this) s
5 313 M
( document are to be interpreted as described in [RFC 2119].) s
5 291 M
(3. Tools) s
5 269 M
( The tools or characteristics that we discuss are the following.) s
5 247 M
(3.1. Characterizing Aggregate Traffic on a Link) s
5 225 M
( o Distribution of per-packet round-trip times.) s
5 203 M
( o Distribution of packet sequence numbers.) s
5 181 M
( o Distribution of packet sizes.) s
5 126 M
(Floyd, Kohler Section 3.1. [Page 4]) s
_R
S
%%Page: (5) 5
%%BeginPageSetup
_S
24 24 translate
/pagenum 5 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( o Ratio between forward-path and reverse-path traffic.) s
5 665 M
( o Distribution of peak flow rates.) s
5 643 M
( o Distribution of transport protocols.) s
5 621 M
(3.2. Characterizing an End-to-End Path) s
5 599 M
( o Synchronization ratio.) s
5 577 M
( o Drop rates as a function of packet size.) s
5 555 M
( o Drop rates as a function of burst size.) s
5 533 M
( o Drop rates as a function of sending rate.) s
5 511 M
( o Degree of packet drops.) s
5 489 M
( o Range of queueing delay.) s
5 467 M
(3.3. Other Characteristics) s
5 445 M
( o Congestion control mechanisms for traffic, along with sender and) s
5 434 M
( receiver buffer sizes.) s
5 412 M
( o Characterization of congested links in terms of bandwidth and) s
5 401 M
( typical levels of congestion \(in terms of packet drop rates\).) s
5 379 M
( o Characterization of congested links in terms of buffer size.) s
5 357 M
( o Characterization of challenging lower layers in terms of) s
5 346 M
( reordering, delay variation, packet corruption, and the like.) s
5 324 M
( o Characterization of network changes affecting congestion, such as) s
5 313 M
( routing changes or link outages.) s
5 291 M
( Below we will discuss each characteristic in turn, giving the) s
5 280 M
( definition, the factors determining that characteristic, the effect) s
5 269 M
( on congestion control metrics, and what is known so far from) s
5 258 M
( measurement studies in the Internet.) s
5 236 M
(4. Distribution of per-packet round-trip times) s
5 214 M
( Definition: The distribution of per-packet round-trip times on a) s
5 203 M
( link is defined formally by assigning to each packet the most recent) s
5 192 M
( round trip time measured for that end-to-end connection. In) s
5 181 M
( practice, coarse-grained information is generally sufficient, even) s
5 170 M
( though it has been shown that there is significant variability in) s
5 126 M
(Floyd, Kohler Section 4. [Page 5]) s
_R
S
%%Page: (6) 6
%%BeginPageSetup
_S
24 24 translate
/pagenum 6 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( round-trip times within a TCP connection [AKSJ03], and it is) s
5 676 M
( sufficient to assign to each packet the first round-trip time) s
5 665 M
( measurement for that connection, or to assign the current round-trip) s
5 654 M
( time estimate maintained by the TCP connection.) s
5 621 M
( Determining factors: The distribution of per-packet round-trip times) s
5 610 M
( on a link is determined by end-to-end propagation delays, by) s
5 599 M
( queueing delays along end-to-end paths, and by the congestion) s
5 588 M
( control mechanisms used by the traffic. For example, for a scenario) s
5 577 M
( using TCP, TCP connections with smaller round-trip times will) s
5 566 M
( receive a proportionally larger fraction of traffic than competing) s
5 555 M
( TCP connections with larger round-trip times, all else being equal,) s
5 544 M
( due to the dynamics of TCP favoring flows with smaller round-trip) s
5 533 M
( times. This will generally shift the distribution of per-packet) s
5 522 M
( RTTs lower relative to the distribution of per-connection RTTs,) s
5 511 M
( since short-RTT connections will have more packets.) s
5 489 M
( Effect on congestion control metrics: The distribution of per-packet) s
5 478 M
( round-trip times on a link affects the burstiness of the aggregate) s
5 467 M
( traffic, and therefore can affect congestion control performance in) s
5 456 M
( a range of areas such as delay/throughput tradeoffs. The) s
5 445 M
( distribution of per-packet round-trip times can also affect metrics) s
5 434 M
( of fairness, degree of oscillations, and the like. For example,) s
5 423 M
( long-term oscillations of queueing delay are more likely to occur in) s
5 412 M
( scenarios with a narrow range of round-trip times [FK02].) s
5 390 M
( Measurements: The distribution of per-packet round-trip times for) s
5 379 M
( TCP traffic on a link can be measured from a packet trace with the) s
5 368 M
( passive TCP round-trip time estimator from Jiang and Dovrolis) s
5 357 M
( [JD02]. [Add pointers to other estimators, such as ones mentioned) s
5 346 M
( in JD02. Add a pointer to Mark Allman's loss detection tool.]) s
5 335 M
( Their paper shows the distribution of per-packet round-trip times) s
5 324 M
( for TCP packets for a number of different links. For the links) s
5 313 M
( measured, the percent of packets with round-trip times at most) s
5 302 M
( 100 ms ranged from 30% to 80%, and the percent of packets with) s
5 291 M
( round-trip times at most 200 ms ranged from 55% to 90%, depending on) s
5 280 M
( the link.) s
5 258 M
( In the NS simulator, the distribution of per-packet round-trip times) s
5 247 M
( for TCP packets on a link can be reported by the queue monitor,) s
5 236 M
( using TCP's estimated round-trip time added to packet headers. This) s
5 225 M
( is illustrated in the validation test "./test-all-simple stats3" in) s
5 214 M
( the directory tcl/test.) s
5 192 M
( Scenarios: [FK02] shows a relatively simple scenario, with a) s
5 181 M
( dumbbell topology with four access links on each end, that gives a) s
5 170 M
( fairly realistic range of round-trip times. [Look for the other) s
5 126 M
(Floyd, Kohler Section 4. [Page 6]) s
_R
S
%%Page: (7) 7
%%BeginPageSetup
_S
24 24 translate
/pagenum 7 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( citations to add.]) s
5 665 M
(5. Distribution of packet sequence numbers) s
5 643 M
( Definition: The distribution of packet sequence numbers on a link is) s
5 632 M
( defined by giving each packet a sequence number, where the first) s
5 621 M
( packet in a connection has sequence number 1, the second packet has) s
5 610 M
( sequence number 2, and so on. The distribution of packet sequence) s
5 599 M
( numbers can be derived in a straightforward manner from the) s
5 588 M
( distribution of connection sizes, and vice versa; however, the) s
5 577 M
( distribution of connection sizes is more suited for traffic) s
5 566 M
( generators, and the distribution of packet sequence numbers is more) s
5 555 M
( suited for measuring and illustrating the packets actually seen on a) s
5 544 M
( link over a fixed interval of time. There has been a considerably) s
5 533 M
( body of research over the last ten years on the heavy-tailed) s
5 522 M
( distribution of connection sizes for traffic on the Internet.) s
5 511 M
( [CBC95] [Add citations.]) s
5 489 M
( Determining factors: The distribution of connection sizes is largely) s
5 478 M
( determined by the traffic generators used in a scenario. For) s
5 467 M
( example, is there a single traffic generator characterized by a) s
5 456 M
( distribution of connection sizes? A mix of long-lived and web) s
5 445 M
( traffic, with the web traffic characterized by a distribution of) s
5 434 M
( connection sizes? Or something else?) s
5 412 M
( Effect on congestion control metrics: The distribution of packet) s
5 401 M
( sequence numbers affects the burstiness of aggregate traffic on a) s
5 390 M
( link, thereby affecting all congestion control metrics for which) s
5 379 M
( this is a factor. As an example, [FK02] illustrates that the) s
5 368 M
( traffic mix can affect the queue dynamics on a congested link.) s
5 357 M
( [Find more to cite, about the effect of the distribution of packet) s
5 346 M
( sequence numbers on congestion control metrics.]) s
5 324 M
( [Add a paragraph about the impact of medium-size flows.]) s
5 302 M
( [Add a paragraph about the impact of flows starting and stopping.]) s
5 280 M
( [Add a warning about scenarios that use only long-lived flows, or a) s
5 269 M
( mix of long-lived and very short flows.]) s
5 247 M
( Measurements: [Cite some of the literature.]) s
5 225 M
( Traffic generators: Some of the available traffic generators are) s
5 214 M
( listed on the web site for "Traffic Generators for Internet Traffic") s
5 203 M
( [TG]. This includes pointers to traffic generators for peer-to-peer) s
5 192 M
( traffic, traffic from online games, and traffic from Distributed) s
5 181 M
( Denial of Service \(DDoS\) attacks.) s
5 126 M
(Floyd, Kohler Section 5. [Page 7]) s
_R
S
%%Page: (8) 8
%%BeginPageSetup
_S
24 24 translate
/pagenum 8 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( In the NS simulator, the distribution of packet sequence numbers for) s
5 676 M
( TCP packets on a link can be reported by the queue monitor at a) s
5 665 M
( router. This is illustrated in the validation test "./test-all-) s
5 654 M
( simple stats3" in the directory tcl/test.) s
5 632 M
(6. The Distribution of Packet Sizes) s
5 610 M
( Definition: The distribution of packet sizes is defined in a) s
5 599 M
( straightforward way, using packet sizes in bytes.) s
5 577 M
( Determining factors: The distribution of packet sizes is determined) s
5 566 M
( by the traffic mix, the path MTUs, and by the packet sizes used by) s
5 555 M
( the transport-level senders.) s
5 533 M
( The distribution of packet sizes on a link is also determined by the) s
5 522 M
( mix of forward-path TCP traffic and reverse-path TCP traffic in that) s
5 511 M
( scenario, for a scenario characterized by a `forward path' \(e.g.,) s
5 500 M
( left to right on a particular link\) and a `reverse path' \(e.g.,) s
5 489 M
( right to left on the same link\). For such a scenario, the forward-) s
5 478 M
( path TCP traffic contributes data packets to the forward link and) s
5 467 M
( acknowledgment packets to the reverse link, while the reverse-path) s
5 456 M
( TCP traffic contributes small acknowledgment packets to the forward) s
5 445 M
( link. The ratio between TCP data and TCP ACK packets on a link can) s
5 434 M
( be used as some indication of the ratio between forward-path and) s
5 423 M
( reverse-path TCP traffic.) s
5 401 M
( Effect on congestion control metrics: The distribution of packet) s
5 390 M
( sizes on a link is an indicator of the ratio of forward-path and) s
5 379 M
( reverse-path TCP traffic in that network. The amount of reverse-) s
5 368 M
( path traffic determines the loss and queueing delay experienced by) s
5 357 M
( acknowledgement packets on the reverse path, significantly affecting) s
5 346 M
( the burstiness of the aggregate traffic on the forward path. [In) s
5 335 M
( what other ways does the distribution of packet sizes affect) s
5 324 M
( congestion control metrics?]) s
5 302 M
( Measurements: There has been a wealth of measurements over time on) s
5 291 M
( the packet size distribution of traffic [A00], [HMTG01]. These) s
5 280 M
( measurements are generally consistent with a model of roughly 10% of) s
5 269 M
( the TCP connections using an MSS of roughly 500 bytes, and with the) s
5 258 M
( other 90% of TCP connections using an MSS of 1460 bytes.) s
5 236 M
(7. The Ratio Between Forward-path and Reverse-path Traffic) s
5 214 M
( Definition: For a scenario characterized by a `forward path' \(e.g.,) s
5 203 M
( left to right on a particular link\) and a `reverse path' \(e.g.,) s
5 192 M
( right to left on the same link\), the ratio between forward-path and) s
5 181 M
( reverse-path traffic can be defined as the ratio between the) s
5 170 M
( forward-path traffic in bps, and the reverse-path traffic in bps.) s
5 126 M
(Floyd, Kohler Section 7. [Page 8]) s
_R
S
%%Page: (9) 9
%%BeginPageSetup
_S
24 24 translate
/pagenum 9 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( Determining factors: The ratio between forward-path and reverse-path) s
5 676 M
( traffic is defined largely by the traffic mix.) s
5 654 M
( Effect on congestion control metrics: Zhang, Shenker and Clark have) s
5 643 M
( shown in 1991 that for TCP, the amount of reverse-path traffic) s
5 632 M
( affects the ACK compression and packet drop rate for TCP) s
5 621 M
( acknowledgement packets, significantly affecting the burstiness of) s
5 610 M
( TCP traffic on the forward path [ZSC91]. The queueing delay on the) s
5 599 M
( reverse path also affects the performance of delay-based congestion) s
5 588 M
( control mechanisms, if the delay is computed based on round-trip) s
5 577 M
( times. This has been shown by Grieco and Mascolo in [GM04] and by) s
5 566 M
( Prasad, Jain, and Dovrolis in [PJD04].) s
5 544 M
( Measurements: There is a need for measurements on the range of) s
5 533 M
( ratios between forward-path and reverse-path traffic for congested) s
5 522 M
( links. In particular, for TCP traffic traversing congested link X,) s
5 511 M
( what is the likelihood that the acknowledgement traffic will) s
5 500 M
( encounter congestion \(i.e., queueing delay, packet drops\) somewhere) s
5 489 M
( on the reverse path as well?) s
5 467 M
( As discussed in Section 6, the distribution of packet sizes on a) s
5 456 M
( link can be used as an indicator of the ratio of forward-path and) s
5 445 M
( reverse-path TCP traffic in that network.) s
5 423 M
(8. The Distribution of Per-Packet Peak Flow Rates) s
5 401 M
( Definition: The distribution of peak flow rates is defined by) s
5 390 M
( assigning to each packet the peak sending rate in bytes per second) s
5 379 M
( of that connection, where the peak sending rate is defined over) s
5 368 M
( 0.1-second intervals. The distribution of peak flow rates gives) s
5 357 M
( some indication of the ratio of "alpha" and "beta" traffic on a) s
5 346 M
( link, where alpha traffic on a congested link is defined as traffic) s
5 335 M
( with that link at the main bottleneck, while the beta traffic on the) s
5 324 M
( link has a primary bottleneck elsewhere along its path [RSB01].) s
5 302 M
( Determining factors: The distribution of peak flow rates is) s
5 291 M
( determined by flows with bottlenecks elsewhere along their end-to-) s
5 280 M
( end path, e.g., flows with low-bandwidth access links. The) s
5 269 M
( distribution of peak flow rates is also affected by applications) s
5 258 M
( with limited sending rates.) s
5 236 M
( Effect on congestion control metrics: The distribution of peak flow) s
5 225 M
( rates affects the burstiness of aggregate traffic, with low-peak-) s
5 214 M
( rate traffic decreasing the aggregate burstiness, and adding to the) s
5 203 M
( traffic's tractability.) s
5 181 M
( Measurements: [RSB01]. The distribution of peak rates can be) s
5 170 M
( expected to change over time, as there is an increasing number of) s
5 126 M
(Floyd, Kohler Section 8. [Page 9]) s
_R
S
%%Page: (10) 10
%%BeginPageSetup
_S
24 24 translate
/pagenum 10 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( high-bandwidth access links to the home, and of high-bandwidth) s
5 676 M
( Ethernet links at work and at other institutions.) s
5 654 M
( Simulators: [For NS, add a pointer to the DelayBox,) s
5 643 M
( "http://dirt.cs.unc.edu/delaybox/", for more easily simulating low-) s
5 632 M
( bandwidth access links for flows.]) s
5 610 M
( Testbeds: In testbeds, Dummynet [Dummynet] and NISTNet [NISTNet]) s
5 599 M
( provide convenient ways to emulate paths with different limited peak) s
5 588 M
( rates.) s
5 566 M
(9. The Distribution of Transport Protocols.) s
5 544 M
( Definition: The distribution of transport protocols on a congested) s
5 533 M
( link is straightforward, with each packet given its associated) s
5 522 M
( transport protocol \(e.g., TCP, UDP\). The distribution is often) s
5 511 M
( given both in terms of packets and in terms of bytes.) s
5 489 M
( For UDP packets, it might be more helpful to classify them in terms) s
5 478 M
( of the port number, or the assumed application \(e.g., DNS, RIP,) s
5 467 M
( games, Windows Media, RealAudio, RealVideo, etc.\) [MAWI]\). Other) s
5 456 M
( traffic includes ICMP, IPSEC, and the like. In the future there) s
5 445 M
( could be traffic from SCTP, DCCP, or from other transport protocols.) s
5 423 M
( Effect on congestion control metrics: The distribution of transport) s
5 412 M
( protocols affects metrics relating to the effectiveness of AQM) s
5 401 M
( mechanisms on a link.) s
5 379 M
( Measurements: In the past, TCP traffic has typically consisted of) s
5 368 M
( 90% to 95% of the bytes on a link [UW02], [UA01]. [Get updated) s
5 357 M
( citations for this.] Measurement studies show that TCP traffic from) s
5 346 M
( web servers almost always uses conformant TCP congestion control) s
5 335 M
( procedures [MAF05].) s
5 313 M
(10. The Synchronization Ratio) s
5 291 M
( Definition: The synchronization ratio is defined as the degree of) s
5 280 M
( synchronization of loss events between two TCP flows on the same) s
5 269 M
( path. Thus, the synchronization ratio is defined as a) s
5 258 M
( characteristic of an end-to-end path. When one TCP flow of a pair) s
5 247 M
( has a loss event, the synchronization ratio is given by the fraction) s
5 236 M
( of those loss events for which the second flow has a loss event) s
5 225 M
( within one round-trip time. Each connection in a flow pair has a) s
5 214 M
( separate synchronization ratio, and the overall synchronization) s
5 203 M
( ratio of the pair of flows is the higher of the two ratios. When) s
5 192 M
( measuring the synchronization ratio, it is preferable to start the) s
5 181 M
( two TCP flows at slightly different times, with large receive) s
5 170 M
( windows.) s
5 126 M
(Floyd, Kohler Section 10. [Page 10]) s
_R
S
%%Page: (11) 11
%%BeginPageSetup
_S
24 24 translate
/pagenum 11 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( Determining factors: The synchronization ratio is determined largely) s
5 676 M
( by the traffic mix on the congested link, and by the AQM mechanism) s
5 665 M
( \(or lack of AQM mechanism\).) s
5 643 M
( Different types of TCP flows are also likely to have different) s
5 632 M
( synchronization measures. E.g., Two HighSpeed TCP flows might have) s
5 621 M
( higher synchronization measures that two Standard TCP flows on the) s
5 610 M
( same path, because of their more aggressive window increase rates.) s
5 599 M
( Raina, Towsley, and Wischik [RTW05] have discussed the relationships) s
5 588 M
( between synchronization and TCP's increase and decrease parameters.) s
5 566 M
( Effect on congestion control metrics: The synchronization ratio) s
5 555 M
( affects convergence times for high-bandwidth TCPs. Convergence) s
5 544 M
( times are known to be poor for some high-bandwidth protocols in) s
5 533 M
( environments with high levels of synchronization. [Cite the papers) s
5 522 M
( by Leith and Shorten.] However, it is not clear if these) s
5 511 M
( environments with high levels of synchronization are realistic.) s
5 489 M
( Wischik and MeKweon [WM05] have shown that the level of) s
5 478 M
( synchronization affects the buffer requirements at congested) s
5 467 M
( routers. Baccelli and Hong [BH02] have a model showing the effect) s
5 456 M
( of the synchronization ratio on aggregate throughput.) s
5 434 M
( Measurements: Grenville Armitage and Qiang Fu have performed initial) s
5 423 M
( experiments of synchronization in the Internet, using Standard TCP) s
5 412 M
( flows, and have found very low levels of synchronization.) s
5 390 M
( In a discussion of the relationship between stability and) s
5 379 M
( desynchronization, Raina, Towsley, and Wischik [RTW05] report that) s
5 368 M
( "synchronization has been reported again and again in simulations".) s
5 357 M
( In contrast, synchronization has not been reported again and again) s
5 346 M
( in the real-world Internet.) s
5 324 M
( Appenzeller, Keslassy, and McKeown in [AKM04] report the following:) s
5 313 M
( "Flows are not synchronized in a backbone router carrying thousands) s
5 302 M
( of flows with varying RTTs. Small variations in RTT or processing) s
5 291 M
( time are sufficient to prevent synchronization [QZK01]; and the) s
5 280 M
( absence of synchronization has been demonstrated in real networks) s
5 269 M
( [F02, IMD01].") s
5 247 M
( [Appenzeller et al, Sizing Router Buffers, reports that) s
5 236 M
( synchronization is rare as the number of competing flows increases.) s
5 225 M
( Kevin Jeffay has some results on synchronization also.]) s
5 203 M
( Needed: We need measurements of the synchronization ratio for flows) s
5 192 M
( that use high-bandwidth protocols over high-bandwidth paths, given) s
5 181 M
( typical levels of competing traffic and with typical queuing) s
5 170 M
( mechanisms at routers \(whatever these are\), to see if there are) s
5 126 M
(Floyd, Kohler Section 10. [Page 11]) s
_R
S
%%Page: (12) 12
%%BeginPageSetup
_S
24 24 translate
/pagenum 12 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( higher levels of synchronization with high-bandwidth protocols such) s
5 676 M
( as HighSpeed TCP, Fast TCP, and the like, which are more aggressive) s
5 665 M
( than Standard TCP. The assumption would be that in many) s
5 654 M
( environments, high-bandwidth protocols have higher levels of) s
5 643 M
( synchronization than flows using Standard TCP.) s
5 621 M
(11. Drop or Mark Rates as a Function of Packet Size) s
5 599 M
( Definition: Drop rates as a function of packet size are defined by) s
5 588 M
( the actual drop rates for different packets on an end-to-end path or) s
5 577 M
( on a congested link over a particular time interval. In some cases,) s
5 566 M
( e.g., Drop-Tail queues in units of packets, general statements can) s
5 555 M
( be made; e.g., that large and small packets will experience the same) s
5 544 M
( packet drop rates. However, in other cases, e.g., Drop-Tail queues) s
5 533 M
( in units of bytes, no such general statement can be made, and the) s
5 522 M
( drop rate as a function of packet size will be determined in part by) s
5 511 M
( the traffic mix at the congested link at that point of time.) s
5 489 M
( Determining factors: The drop rate as a function of packet size is) s
5 478 M
( determined in part by the queue architecture. E.g., is the Drop-) s
5 467 M
( Tail queue in units of packets, of bytes, of 60-byte buffers, or of) s
5 456 M
( a mix of buffer sizes? Is the AQM mechanism in packet mode,) s
5 445 M
( dropping each packet with the same probability, or in byte mode,) s
5 434 M
( with the probability of dropping or marking a packet being) s
5 423 M
( proportional to the packet size in bytes.) s
5 401 M
( The effect of packet size on drop rate would also be affected by the) s
5 390 M
( presence of preferential scheduling for small packets, or by) s
5 379 M
( differential scheduling for packets from different flows \(e.g., per-) s
5 368 M
( flow scheduling, or differential scheduling for UDP and TCP) s
5 357 M
( traffic\).) s
5 335 M
( In many environments, the drop rate as a function of packet size) s
5 324 M
( will be heavily affected by the traffic mix at a particular time.) s
5 313 M
( For example, is the traffic mix dominated by large packets, or by) s
5 302 M
( smaller ones? In some cases, the overall packet drop rate could) s
5 291 M
( also affect the relative drop rates for different packet sizes.) s
5 269 M
( In wireless networks, the drop rate as a function of packet size is) s
5 258 M
( also determined by the packet corruption rate as a function of) s
5 247 M
( packet size. [Cite Deborah Pinck's papers on Satellite-Enhanced) s
5 236 M
( Personal Communications Experiments and on Experimental Results from) s
5 225 M
( Internetworking Data Applications Over Various Wireless Networks) s
5 214 M
( Using a Single Flexible Error Control Protocol.] [Cite the general) s
5 203 M
( literature.]) s
5 181 M
( Effect on congestion control metrics: The drop rate as a function of) s
5 170 M
( packet size has a significant effect on the performance of) s
5 126 M
(Floyd, Kohler Section 11. [Page 12]) s
_R
S
%%Page: (13) 13
%%BeginPageSetup
_S
24 24 translate
/pagenum 13 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( congestion control for VoIP and other small-packet flows.) s
5 676 M
( [Citation: "TFRC for Voice: the VoIP Variant", draft-ietf-dccp-tfrc-) s
5 665 M
( voip-02.txt, and earlier papers.] The drop rate as a function of) s
5 654 M
( packet size also has an effect on TCP performance, as it affects the) s
5 643 M
( drop rates for TCP's SYN and ACK packets. [Citation: Jeffay and) s
5 632 M
( others.]) s
5 610 M
( Measurements: We need measurements of the drop rate as a function of) s
5 599 M
( packet size over a wide range of paths, or for a wide range of) s
5 588 M
( congested links. For tests of relative drop rates on end-to-end) s
5 577 M
( packets, one possibility would be to run successive TCP connections) s
5 566 M
( with 200-byte, 512-byte, and 1460-byte packets, and to compare the) s
5 555 M
( packet drop rates. The ideal test would include running TCP) s
5 544 M
( connections on the reverse path, to measure the drop rates for the) s
5 533 M
( small ACK packets on the forward path. It would also be useful to) s
5 522 M
( characterize the difference in drop rates for 200-byte TCP packets) s
5 511 M
( and 200-byte UDP packets, even though some of this difference could) s
5 500 M
( be due to the relative burstiness of the different connections.) s
5 478 M
( Ping experiments could also be used to get measurements of drop) s
5 467 M
( rates as a function size, but it would be necessary to make sure) s
5 456 M
( that the ping sending rates were adjusted to be TCP-friendly.) s
5 434 M
( [Cite the known literature on drop rates as a function of packet) s
5 423 M
( size.]) s
5 401 M
( Our conjecture is that there is a wide range of behaviors for this) s
5 390 M
( characteristic in the real world. Routers include Drop-Tail queues) s
5 379 M
( in packets, bytes, and buffer sizes in between; these will have) s
5 368 M
( quite different drop rates as a function of packet size. Some) s
5 357 M
( routers include RED in byte mode \(the default for RED in Linux\) and) s
5 346 M
( some have RED in packet mode \(Cisco, I believe\). This also affects) s
5 335 M
( drop rates as a function of packet size.) s
5 313 M
( Some routers on congested access links use per-flow scheduling. In) s
5 302 M
( this case, does the per-flow scheduling have the goal of fairness in) s
5 291 M
( *bytes* per second or in *packets* per second? What effect does the) s
5 280 M
( per-flow scheduling have on the drop rate as a function of packet) s
5 269 M
( size, for packets in different flows \(e.g., a small-packet VoIP flow) s
5 258 M
( competing against a large-packet TCP flow\) or for packets within the) s
5 247 M
( same flow \(small ACK packets and large data packets on a two-way TCP) s
5 236 M
( connection\).) s
5 214 M
(12. Drop Rates as a Function of Burst Size.) s
5 192 M
( Definition: Burst-tolerance, or drop rates as a function of burst) s
5 181 M
( size, can be defined in terms of an end-to-end path, or in terms of) s
5 170 M
( aggregate traffic on a congested link.) s
5 126 M
(Floyd, Kohler Section 12. [Page 13]) s
_R
S
%%Page: (14) 14
%%BeginPageSetup
_S
24 24 translate
/pagenum 14 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( The burst-tolerance of an end-to-end path is defined in terms of) s
5 676 M
( connections with different degrees of burstiness within a round-trip) s
5 665 M
( time. When packets are sent in bursts of N packets, does the drop) s
5 654 M
( rate vary as a function of N? For example, if the TCP sender sends) s
5 643 M
( small bursts of K packets, for K less than the congestion window,) s
5 632 M
( how does the size of K affect the loss rate? Similarly, for a ping) s
5 621 M
( tool sending pings at a certain rate in packets per second, one) s
5 610 M
( could see how the clustering of the ping packets in clusters of size) s
5 599 M
( K affects the packet drop rate. As always with such ping) s
5 588 M
( experiments, it would be important to adjust the sending rate to) s
5 577 M
( maintain a longer-term sending rate that was TCP-friendly.) s
5 555 M
( Determining factors: The burst-tolerance is determined largely by) s
5 544 M
( the AQM mechanisms for the congested routers on a path, and by the) s
5 533 M
( traffic mix. For a Drop-Tail queue with only a small number of) s
5 522 M
( competing flows, the burst-tolerance is likely to be low, and for) s
5 511 M
( AQM mechanisms where the packet drop rate is a function of the) s
5 500 M
( average queue size rather than the instantaneous queue size, the) s
5 489 M
( burst tolerance should be quite high.) s
5 467 M
( Effect on congestion control metrics: The burst-tolerance of the) s
5 456 M
( path or congested link can affect fairness between competing flows) s
5 445 M
( with different round-trip times; for example, Standard TCP flows) s
5 434 M
( with longer round-trip times are likely to have a more bursty) s
5 423 M
( arrival pattern at the congested link that that of Standard TCP) s
5 412 M
( flows with shorter round-trip times. As a result, in environment) s
5 401 M
( with low burst tolerance \(e.g., scenarios with Drop-Tail queues\),) s
5 390 M
( longer-round-trip-time TCP connections can see higher packet drop) s
5 379 M
( rates than other TCP connections, and receive an even smaller) s
5 368 M
( fraction of the link bandwidth than they would otherwise.) s
5 357 M
( [FJ92]\(Section 3.2\). We note that some TCP traffic is inherently) s
5 346 M
( bursty, e.g., Standard TCP without rate-based pacing, particularly) s
5 335 M
( in the presence of dropped ACK packets or of ACK compression. The) s
5 324 M
( burst-tolerance of a router can also affect the delay-throughput) s
5 313 M
( tradeoffs and packet drop rates of the path or of the congested) s
5 302 M
( link.) s
5 280 M
( Measurements: One could measure the burst-tolerance of an end-to-end) s
5 269 M
( path by running successive TCP connections, forcing bursts of size) s
5 258 M
( at least K by dropping an appropriate fraction of the ACK packets to) s
5 247 M
( the TCP receiver. Alternately, if one had control of the TCP) s
5 236 M
( sender, one could modify the TCP sender to send bursts of K packets) s
5 225 M
( when the congestion window was K or more segments.) s
5 203 M
( [Look at Crovella's paper on loss pairs.]) s
5 181 M
( [Look at: M. Allman and E. Blanton, "Notes on Burst Mitigation for) s
5 170 M
( Transport Protocols", ACM Computer Communication Review, vol. 35\(2\),) s
5 126 M
(Floyd, Kohler Section 12. [Page 14]) s
_R
S
%%Page: (15) 15
%%BeginPageSetup
_S
24 24 translate
/pagenum 15 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( \(2005\).]) s
5 665 M
( Making inferences about the AQM mechanism for the congested router) s
5 654 M
( on an end-to-end path: One potential use of measurement tools for) s
5 643 M
( determining the burst-tolerance of an end-to-end path would be to) s
5 632 M
( make inferences about the presence or absence of an AQM mechanism at) s
5 621 M
( the congested link or links. As a simple test, one could run a TCP) s
5 610 M
( connection until the connection comes out of slow-start. If the) s
5 599 M
( receive window of the TCP connection was sufficiently high that the) s
5 588 M
( connection exited slow-start with packet drops or marks instead of) s
5 577 M
( because of the limitation of the receive window, one could record) s
5 566 M
( the congestion window at the end of slow-start, and the number of) s
5 555 M
( packets dropped from this window. A high packet drop rate might be) s
5 544 M
( more typical of a Drop-Tail queue with small-scale statistical) s
5 533 M
( multiplexing on the congested link, and a single packet drop coming) s
5 522 M
( out of slow-start would suggest an AQM mechanism at the congested) s
5 511 M
( link.) s
5 489 M
( The synchronization measure could also add information about the) s
5 478 M
( likely presence or absence of AQM on the congested link\(s\) of an) s
5 467 M
( end-to-end path, with paths with higher levels of synchronization) s
5 456 M
( being more likely to have Drop-Tail queues with small-scale) s
5 445 M
( statistical multiplexing on the congested link\(s\).) s
5 423 M
( [Cite the relevant literature about tools for determining the AQM) s
5 412 M
( mechanism on an end-to-end path.]) s
5 390 M
(13. Drop Rates as a Function of Sending Rate.) s
5 368 M
( Definition: Drop rates as a function of sending rate is defined in) s
5 357 M
( terms of the drop behavior of a flow in the end-to-end path. That) s
5 346 M
( is, does the sending rate of an individual flow affect its own) s
5 335 M
( packet drop rate, or its packet drop rate largely independent of the) s
5 324 M
( sending rate of the flow?) s
5 302 M
( Determining factors: The sending rate of the flow affects its own) s
5 291 M
( packet drop rate in an environment with small-scale statistical) s
5 280 M
( multiplexing on the congested link. The packet drop rate is largely) s
5 269 M
( independent of the sending rate in an environment with large-scale) s
5 258 M
( statistical multiplexing, with many competing small flows at the) s
5 247 M
( congested link. Thus, the behavior of drop rates as a function of) s
5 236 M
( sending rate is a rough measure of the level of statistical) s
5 225 M
( multiplexing on the congested links of an end-to-end path.) s
5 203 M
( Effect on congestion control metrics: The level of statistical) s
5 192 M
( multiplexing at the congested link can affect the performance of) s
5 181 M
( congestion control mechanisms in transport protocols. For example,) s
5 170 M
( delay-based congestion control is often better suited to) s
5 126 M
(Floyd, Kohler Section 13. [Page 15]) s
_R
S
%%Page: (16) 16
%%BeginPageSetup
_S
24 24 translate
/pagenum 16 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( environments small-scale statistical multiplexing at the congested) s
5 676 M
( link, where the transport protocol responds to the delay caused by) s
5 665 M
( its own sending rate.) s
5 643 M
( Measurements: In a simulation or testbed, the level of statistical) s
5 632 M
( multiplexing on the congested link can be observed directly. In the) s
5 621 M
( Internet, the level of statistical multiplexing on the congested) s
5 610 M
( links of an end-to-end path can be inferred indirectly through per-) s
5 599 M
( flow measurements, by observing whether the packet drop rate varies) s
5 588 M
( as a function of the sending rate of the flow.) s
5 566 M
(14. Congestion Control Mechanisms for Traffic, along with Sender and) s
5 555 M
( Receiver Buffer Sizes.") s
5 533 M
( Effect on congestion control metrics: Please don't evaluate AQM) s
5 522 M
( mechanisms by using Reno TCP, or evaluate new transport protocols by) s
5 511 M
( comparing them with the performance of Reno TCP! For an) s
5 500 M
( explanation, see [FK02]\(Section 3.4\).) s
5 478 M
( Measurements: See [MAF05].) s
5 456 M
(15. Characterization of Congested Links in Terms of Bandwidth and) s
5 445 M
(Typical Levels of Congestion) s
5 423 M
( [Pointers to the current state of our knowledge.]) s
5 401 M
(16. Characterization of Challenging Lower Layers.) s
5 379 M
( [This will just be a short set of pointers to the relevant) s
5 368 M
( literature, which is quite extensive.]) s
5 346 M
(17. Characterization of Network Changes Affecting Congestion) s
5 324 M
( [Pointers to the current state of our knowledge.]) s
5 302 M
(18. Using the Tools Presented in this Document) s
5 280 M
( [To be done.]) s
5 258 M
(19. Related Work) s
5 236 M
( [Cite "On the Effective Evaluation of TCP" by Allman and Falk.]) s
5 214 M
(20. Conclusions) s
5 192 M
( [To be done.]) s
5 126 M
(Floyd, Kohler Section 20. [Page 16]) s
_R
S
%%Page: (17) 17
%%BeginPageSetup
_S
24 24 translate
/pagenum 17 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
(21. Security Considerations) s
5 665 M
( There are no security considerations in this document.) s
5 643 M
(22. IANA Considerations) s
5 621 M
( There are no IANA considerations in this document.) s
5 599 M
(23. Acknowledgements) s
5 577 M
( Thanks to Xiaoliang \(David\) Wei for feedback and contributions to) s
5 566 M
( this document.) s
5 544 M
(Informative References) s
5 522 M
( [RFC 2119] S. Bradner. Key Words For Use in RFCs to Indicate) s
5 511 M
( Requirement Levels. RFC 2119.) s
5 489 M
( [MAWI] M.W. Group, Mawi working group traffic archive, URL) s
5 478 M
( "http://tracer.csl.sony.jp/mawi/".) s
5 456 M
( [AKM04] B. Appenzeller, I. Keslassy, and N. McKeown, Sizing Router) s
5 445 M
( Buffers, SIGCOMM 2004.) s
5 423 M
( [AKSJ03] J. Aikat, J. Kaur, F.D. Smith, and K. Jeffay, Variability) s
5 412 M
( in TCP Roundtrip Times, ACM SIGCOMM Internet Measurement) s
5 401 M
( Conference, Maimi, FL, October 2003, pp. 279-284.) s
5 379 M
( [A00] M. Allman, A Web Server's View of the Transport Layer,) s
5 368 M
( Computer Communication Review, 30\(5\), October 200.) s
5 346 M
( [BH02] F. Baccelli and D. Hong, AIMD, Fairness and Fractal Scaling) s
5 335 M
( of TCP Traffic, Infocom 2002.) s
5 313 M
( [CBC95] C. Cunha, A. Bestavros, and M. Crovella, "Characteristics of) s
5 302 M
( WWW Client-based Traces", BU Technical Report BUCS-95-010, 1995.) s
5 280 M
( [Dummynet] L. Rizzo, Dummynet, URL) s
5 269 M
( "http://info.iet.unipi.it/~luigi/ip_dummynet/".) s
5 247 M
( BIBREF F02 "F02" C. J. Fraleigh, Provisioning Internet Backbone) s
5 236 M
( Networks to Support Latency Sensitive Applications. PhD thesis,) s
5 225 M
( Stanford University, Department of Electrical Engineering, June) s
5 214 M
( 2002.) s
5 192 M
( [FJ92] S. Floyd and V. Jacobson, On Traffic Phase Effects in Packet-) s
5 181 M
( Switched Gateways, Internetworking: Research and Experience, V.3) s
5 170 M
( N.3, September 1992, p.115-156.) s
5 126 M
(Floyd, Kohler [Page 17]) s
_R
S
%%Page: (18) 18
%%BeginPageSetup
_S
24 24 translate
/pagenum 18 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( [FK02] S. Floyd and E. Kohler, Internet Research Needs Better) s
5 676 M
( Models, Hotnets-I, October 2002.) s
5 654 M
( [GM04] L. Grieco and S. Mascolo, Performance Evaluation and) s
5 643 M
( Comparison of Westwood+, New Reno, and Vegas TCP Congestion) s
5 632 M
( Control, CCR, April 2004.) s
5 610 M
( [HMTG01] C. Hollot, V. Misra, D. Towsley, and W. Gong, On Designing) s
5 599 M
( Improved Controllers for AQM Routers Supporting TCP Flows, IEEE) s
5 588 M
( Infocom, 2001.) s
5 566 M
( [IMD01] G. Iannaccone, M. May, and C. Diot, Aggregate Traffic) s
5 555 M
( Performance with Active Queue Management and Drop From Tail.) s
5 544 M
( SIGCOMM Comput. Commun. Rev., 31\(3\):4-13, 2001.) s
5 522 M
( [JD02] H. Jiang and C. Dovrolis, Passive Estimation of TCP Round-) s
5 511 M
( trip Times, Computer Communication Review, 32\(3\), July 2002.) s
5 489 M
( [MAF05] A. Medina, M. Allman, and A. Floyd. Measuring the Evolution) s
5 478 M
( of Transport Protocols in the Internet. Computer Communication) s
5 467 M
( Review, April 2005.) s
5 445 M
( [NISTNet] NIST Net, URL "http://snad.ncsl.nist.gov/itg/nistnet/".) s
5 423 M
( [PJD04] R. Prasad, M. Jain, and C. Dovrolis, On the Effectiveness of) s
5 412 M
( Delay-Based Congestion Avoidance, PFLDnet 2004, February 2004.) s
5 379 M
( [1] L. Qiu, Y. Zhang, and S. Keshav, Understanding the Performance) s
5 368 M
( of Many TCP Flows, Comput. Networks, 37\(3-4\):277-306, 2001.) s
5 346 M
( [RSB01] R. Riedi, S. Sarvotham, and R. Varaniuk, Connection-level) s
5 335 M
( Analysis and Modeling of Network Traffic, SIGCOMM Internet) s
5 324 M
( Measurement Workshop, 2001.) s
5 291 M
( [RTW05] G. Raina, D. Towsley, and D. Wischik, Control Theory for) s
5 280 M
( Buffer Sizing, CCR, July 2005.) s
5 258 M
( [TG] Traffic Generators for Internet Traffic Web Page, URL) s
5 247 M
( "http://www.icir.org/models/trafficgenerators.html".) s
5 225 M
( [UA01] U. of Auckland, Auckland-vi trace data, June 2001. URL) s
5 214 M
( "http://wans.cs.waikato.ac.nz/wand/wits/auck/6/".) s
5 192 M
( [UW02] UW-Madison, Network Performance Statistics, October 2002.) s
5 181 M
( URL "http://wwwstats.net.wisc.edu/".) s
5 126 M
(Floyd, Kohler [Page 18]) s
_R
S
%%Page: (19) 19
%%BeginPageSetup
_S
24 24 translate
/pagenum 19 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( [WM05] D. Wischik and N. McKeown, Buffer sizes for Core Routers,) s
5 676 M
( CCR, July 2005.) s
5 654 M
( [ZCS91] L. Zhang, S. Shenker, and D.D. Clark, Observations and) s
5 643 M
( Dynamics of a Congestion Control Algorithm: the Effects of Two-) s
5 632 M
( way Traffic, SIGCOMM 1991.) s
5 599 M
(Editors' Addresses) s
5 577 M
( Sally Floyd <floyd@icir.org>) s
5 566 M
( ICSI Center for Internet Research) s
5 555 M
( 1947 Center Street, Suite 600) s
5 544 M
( Berkeley, CA 94704) s
5 533 M
( USA) s
5 511 M
( Eddie Kohler <kohler@cs.ucla.edu>) s
5 500 M
( 4531C Boelter Hall) s
5 489 M
( UCLA Computer Science Department) s
5 478 M
( Los Angeles, CA 90095) s
5 467 M
( USA) s
5 445 M
(Full Copyright Statement) s
5 423 M
( Copyright \(C\) The Internet Society 2005. This document is subject) s
5 412 M
( to the rights, licenses and restrictions contained in BCP 78, and) s
5 401 M
( except as set forth therein, the authors retain all their rights.) s
5 379 M
( This document and the information contained herein are provided on) s
5 368 M
( an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE) s
5 357 M
( REPRESENTS OR IS SPONSORED BY \(IF ANY\), THE INTERNET SOCIETY AND THE) s
5 346 M
( INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR) s
5 335 M
( IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF) s
5 324 M
( THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED) s
5 313 M
( WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.) s
5 291 M
(Intellectual Property) s
5 269 M
( The IETF takes no position regarding the validity or scope of any) s
5 258 M
( Intellectual Property Rights or other rights that might be claimed) s
5 247 M
( to pertain to the implementation or use of the technology described) s
5 236 M
( in this document or the extent to which any license under such) s
5 225 M
( rights might or might not be available; nor does it represent that) s
5 214 M
( it has made any independent effort to identify any such rights.) s
5 203 M
( Information on the procedures with respect to rights in RFC) s
5 192 M
( documents can be found in BCP 78 and BCP 79.) s
5 126 M
(Floyd, Kohler [Page 19]) s
_R
S
%%Page: (20) 20
%%BeginPageSetup
_S
24 24 translate
/pagenum 20 def
/fname (draft-irtf-tmrg-tools-01.txt) def
/fdir () def
/ftail (draft-irtf-tmrg-tools-01.txt) def
/user_header_p false def
%%EndPageSetup
5 720 M
(INTERNET-DRAFT Expires: April 2006 October 2005) s
5 687 M
( Copies of IPR disclosures made to the IETF Secretariat and any) s
5 676 M
( assurances of licenses to be made available, or the result of an) s
5 665 M
( attempt made to obtain a general license or permission for the use) s
5 654 M
( of such proprietary rights by implementers or users of this) s
5 643 M
( specification can be obtained from the IETF on-line IPR repository) s
5 632 M
( at http://www.ietf.org/ipr.) s
5 610 M
( The IETF invites any interested party to bring to its attention any) s
5 599 M
( copyrights, patents or patent applications, or other proprietary) s
5 588 M
( rights that may cover technology that may be required to implement) s
5 577 M
( this standard. Please address the information to the IETF at ietf-) s
5 566 M
( ipr@ietf.org.) s
5 126 M
(Floyd, Kohler [Page 20]) s
_R
S
%%Trailer
%%Pages: 20
%%DocumentNeededResources: font Courier-Bold Courier
%%EOF
| PAFTECH AB 2003-2026 | 2026-04-23 14:27:29 |