Feb 2, 2002 - out-degree, different points in this tradeoff can be selected. ..... We used the Georgia Tech "Transit-Stub" model (GT-ITM [19]) to generate the ...
! " #$%#&'()&* + ' *('*('*(
!" #" $ % & ' ()*+, - ' ' &
Æ
! " #! $ % & "! ' ( ) * +! % Æ , - . , '
Æ / ! 0 $ .
1 2 '3! 2 &! 1 4 ''! 52 '! 5 (! . *! 6 +! .
Æ
. , 0 7 8 2 7 2
Æ
9 7 52 5 Æ
: ; 7 8 52 5 / 7 52 5 < 8 =
7 = #
= & 8 1 8 = " = ) / = + =
'3
(
Fully Connected
Bad Overlay
Good Overlay
13 A
A
C
10 r1
r2
2
1
B
2
A
B
D
12
D B
r1
(a)
C
2
3 12
D
12
B
D
(c1)
(b1)
C
A
13
3
13
1
1
C 12
12
r2
(d1)
r1
r1
r2
Physical Topology
r2
A A B
C (b2)
D
A
D B
C (c2)
D B
C (d2)
> ',
!
Æ
% % $ 8 '33
/ ? Æ = 8 & '( " ' !
1 0
1@6 0 1@6 8
8
= 8 . 0
Æ
2 , -
> Æ -
':; 8 > ': ';
1@6 ' A , > ': (; 5 @
8 > ': '; <
1' 1( 1@6
: )B' :5; (#B( :; (&B( :5@;; <
52 5 C 5 : 2; > ':'; . 1@6D
. 2 , Æ
,
Æ
#
.
% 7
?
< 7 :> (;
< 7
8
: 8 ; . , $ % A Æ
%
> (, " #
. ,
<
: ; 5
E &
: 8 ''!; /
.
< 7 / 2 &! < 2
E 8 & '#! . 8
8
> 2 &!
. E , 0
8 F
8 $ 5 : ;
.
/ E
: ; .
0 E .
E : E
"
;
0
?
A > 8
( D < 8 &3G
'!
<
: ; 0
E /
E
,
$ .
.
$
?
>
% /
.
.
A
)
$
5 %
% .
%
9
. 7
0 52 '! 5 (! -
C '& '"!
! " #
< 52
5
(
=
5
/ 52 , . '! 5
28 % 8
'!
8 8 = 52
%
! < .
/ *
< 5
0 > 8 H (
: H : ;
$; ( ( ( ( .
I ( I ( I ( / 5 , /
? > 8
3
) 8 # . . 8 : ;
. ')! <
8
! .
$ %
<
.
0 '*!
, " >
! .
+
!
" # $
. 7 Æ .
? / 0
$
E 8 D / C . J. = K :C.
"
'
$(
> +3 1@6 7 2 % $ /
% 1@6 2 7 : (3G; $ '333
0 1@6 '3(#
7 & 1@6 '3(#
2 '
1@6 $ 7 2 7 1@6 (+ & $ '(*
'3(#
/ % (3G % 1@6 0 2 %
%
Æ
''
6
Kudos 90th percentile Narada 90th percentile Kudos 50th percentile Narada 50th percentile
5
RDP
4 3 2 1 0 100
1000 Overlay size
> , $% &!' ( ) * ( + 30
Kudos Narada
Worst Link Stress
25 20 15 10 5 0 100
1000 Overlay size
> #, ,- . ( * ) ( + '
%!
< > # $ 2 7 . % 8 0 :
; / $ 7 '( '* $ '(*
'3(#
!"# $
'(
N
7 2 # '3 '(3 * '& ('# '" () (+& ( 3 #( "# ' )#& '(* ''*( . ', . 8 7 2 $ $ '
) ( * + +
.
7 2 < . ' (&"
J K 0
E % . '"
'"
'" %
E . , 7 8 ?
2
. 7
<
2 '
( ) * ) + ,
. % 52 5 1@6 . J K : ; 1@6 1@6 A %
'
CAN Best CAN Smart CAN Naive
14
90th percentile RDP
12 10 8 6 4 2 0
100
1000 Overlay size
> &, ( ( * &!' /" ( + 140
Worst Link Stress
120
CAN Best CAN Smart CAN Naive
100 80 60 40 20 0 100
1000 Overlay size
> ", ( ( * ,
< > & 1@6 52 52
" 7 . % $ 1@6 8 J K 0 J K J K % . 52
$ 8 . '#
14
90th percentile RDP
12
Chord Best Chord Smart Chord Naive
10 8 6 4 2 0 100
1000 Overlay size
> ), ( ( &!' ( + < > " $ 52 > 52 ,
> ) +3 1@6 5 = 52
> *
$ % = / J K 5 % 52 A 52 5 -
)#
< 7 52 5 < 8
7 52 " > 52 5 J K -
'&
140
Worst Link Stress
120
Chord Best Chord Smart Chord Naive
100 80 60 40 20 0 100
1000 Overlay size
> *, ( ( 14
90th percentile RDP
12
Kudos Chord Smart CAN Smart Brite Flooding
10 8 6 4 2 0 100
1000 Overlay size
> +, $% &!' ( ( ( + -
$(
> + +3 1@6 % $
> 7 1@6 = 5 52 5 $ -
< > '3 $ % / 7 > 7 $ '"
Worst Link Stress
200
150
Kudos Chord Smart CAN Smart Brite Flooding
100
50
0 100
1000 Overlay size
> '3, - ( ( + (
90th percentile RDP
20
Kudos CAN Brite
15
10
5
0 0
2
4
6 8 Average Degree
10
12
14
> '', ( $% &!' 7 D
1 0 -
-
./ 0 (
28 8 %
<
1@6 > '3(#
52 7 /
5
5
> '' 8 8 1@6 52
# " ')
Worst Access Link Stress
14
Kudos CAN Brite
12 10 8 6 4 2 0 0
5
10
15
20
90th percentile RDP
> '(, &!' " ( . 7 7 8 1@6 < > '( ' 8 1@6
> '( : ; > ' : ; % 1@6 A %
0
, 7 -
?
2
8 > '*
Worst Backbone link stress
70
Kudos CAN Brite
60 50 40 30 20 10 0
5
10 90th percentile RDP
15
20
> ' , &!' " ( Æ
: 52 ; < 2
8 : ; < ( $ : ; 5
: ; > 52 5
A 52 8 :( ; 5 :: ;; > -
. ( $ 5
52 5 2 7 7 2 .
+ /0
/ : 8 & ' ( " * + (3 '(!; L
&'
%
'+
6
: ;
2 7 52 : ; ( 5 : ; 1 6 L : ; . (,
%
/ 0
< = % ('! ((! ('!
7
$ / 7 ('! < ((! $ < 0 0 A
7
1
) 2 /0
. , 0 7 8
2 9 7 2 7 2 ?
= 0 7 52 5
/ % :1@6;
% (3
% / 7 0
< 8
/ 8 7 Æ +
'! = 1 6 > A 1 7 = = J = 5 2 K )*+ ,-*.++ (33' (! < = 1 @ 7 > 7 A J5 ,
K )*+ ,-*.++ (33' ! A J , . K * ) &#O"3
'++#
)*+
#! J" / 6K ,BB" B &! P 5 = E 1 A Q J 5 = K )*+ ,* R (333
,-+/
"! P 5 = 5
J < 5 @
< = K > (333 )! JC K ,BB *! P Q R @ 7 $ @ R J. , < > / L 1 K . 1 95B5=@ 3' ''#' 95 (33' +! 1 6 @ J6 , = @ E L 1 L = 6 6 = K ,",0)*+ , * + 2 (33' ('
'3! @ C A > 7 1 J. 5 1 2 K 1 . 2,,, (33' ''! P 5 = 5
J1 4, 1 A 2 K , ".*.+ 3 (333 )+&O*3# '(! R R
@ 7 C% 7 L R > 7 R / D.
J , 1 2 K . , (333 ' ! @ 6 . JL
