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 2003-20262026-04-23 14:27:29