...

大容量フォトニックネットワークのアーキテクチャ

by user

on
Category: Documents
4

views

Report

Comments

Transcript

大容量フォトニックネットワークのアーキテクチャ
一般社団法人 電子情報通信学会
THE INSTITUTE OF ELECTRONICS,
INFORMATION AND COMMUNICATION ENGINEERS
信学技報
IEICE Technical Report
大容量フォトニックネットワークのアーキテクチャ
長谷川 浩
佐藤 健一
名古屋大学 大学院工学研究科 電子情報システム専攻
〒464-8603 愛知県 名古屋市 千種区 不老町
E-mail: {hasegawa,sato}@nuee.nagoya-u.ac.jp
あらまし 本稿では、超大容量通信を実現するためのフォトニックネットワークのアーキテクチャについて解説
する。従来の IP 技術をベースにしたネットワークでは、パケット単位での経路制御を電気処理にて行うため、宛先
検索にかかるオーバーヘッドが通信容量を制限しかつ膨大な消費電力に直結していた。光ファイバ中の波長多重信
号の経路制御を、波長をラベルとし光信号のまま行うことにより、超低消費電力と大容量を実現するフォトニック
ネットワークが導入されつつあるが、コスト面やハードウェア規模の面から多数の光信号の経路制御処理は容易で
はなく、更なる大容量化は困難であった。本稿ではフォトニックネットワークの現状と課題について述べ、ボトル
ネックであるフォトニックノードの容量を向上させるためのアーキテクチャを解説する。
キーワード
チャ
フォトニックネットワーク 階層化光パス エラスティック光パス フォトニックノードアーキテク
Architectures of Bandwidth Abundant Photonic Networks
Hiroshi HASEGAWA
Ken-ichi SATO
Dept. Electrical Engineering and Computer Science, Nagoya University
Furo-cho, Chikusa-ku, Nagoya 464-8603 Japan
E-mail: {hasegawa, sato}@nuee.nagoya-u.ac.jp
Abstract In this manuscript, we present architectures of photonic networks to realize bandwidth abundant optical transport.
Current IP-based networks are suffered from the energy consumption and the capacity limitation caused by the
packet-by-packet forwarding in the electrical layer. This fact motivates the introduction of photonic networks that utilize
wavelength routing in the optical layer. The elimination of costly E/O and O/E conversion and the routing in the optical layer
makes photonic networks energy efficient and capacity abundant. However, due to the difficulty in realizing large scale optical
switches, further capacity enhancement is not straightforward. We elucidate the current situation and issues in photonic
networks and then show novel node architectures that can achieve larger capacity cost-effectively.
Keyword Photonic Network, Hierarchical Optical Path, Elastic Optical Path, Photonic Node Architecture
This article is a technical report without peer review, and its polished and/or extended version may be published elsewhere.
Copyright ©2013
by IEICE
⫼ᬒ –ቑຍ䛧⥆䛡䜛㏻ಙ䝖䝷䝣䜱䝑䜽ୗ䜚䝖䝷䝣䜱䝑䜽
1.696Tbps (2011/11)
ୖ䜚䝖䝷䝣䜱䝑䜽
0.669Tbps (2011/11)
1998
2000
2002
2004
2006
2008
2010
2012
JPIX (Japan Internet Exchange)䛷䛾䝖䝷䝣䜱䝑䜽ኚື(1998-2012)
90,000
VoIP
Online gaming
80,000
70,000
⫼ᬒ
2003
2004
2005
2006
2007
2008
2009
2010
2011
᪥ᮏᅜෆ䛾⥲㏻ಙ㔞(⥲ົ┬Ⓨ⾲㈨ᩱ䜘䜚)
PB/Month
60,000
• 䝖䝷䝣䜱䝑䜽䛿౫↛䛸䛧䛶ቑຍ
(+30-40%/ᖺ).
50,000
40,000
• ᫎീ୰ᚰ䛾䝃䞊䝡䝇䛜ቑຍ
Web/e-mail/data
30,000
20,000
Internet Video
• ྛ✀⛣ື➃ᮎ䛾ቑຍ
10,000
File Sharing
0
2011 2012 2013 2014 2015 2016
Cisco VNI, “Global Consumer Internet Traffic, 2011-2016”
3
䝕䜱䝇䝥䝺䜲䛾㧗⢭⣽໬
㏻ಙ⏘ᴗ䛾Ⓨᒎ䠄᪥ᮏ’10ї’11䠅
䝇䝬䞊䝖䝣䜷䞁
ಖ᭷⋡
䜽䝷䜴䝗䝃䞊䝡䝇
฼⏝⋡ (௻ᴗ)
ᵑᵌᵎಸ
ᵏᵌᵓಸ
9.7ї29.3%
14.1ї21.6%
䝖䝷䝣䜱䝑䜽䛾ᛶ㉁䛾ኚ໬
䝕䞊䝍㻌 㻌 㻌 㻌 㻌 㻌 㻌 㻌 㻌 㻌 㻌 㻌 ᫎീ୺య䜈
9䝤䝻䞊䝗䝞䞁䝗䜰䜽䝉䝇ᢏ⾡䛾㐍ᒎ (ADSL, FTTH)
ᫎീᢏ⾡䛾㐍ᒎ
iPhone3 : 480 x 320
iPhone4 : 960 x 640
(DVD : 720 x 480)
HDTV: 1920 x 1080
2560 x 1600 (6.1inch)
䜲䞁䝍䞊䝛䝑䝖
᥋⥆ྍ⬟
䝔䝺䝡ಖ᭷⋡
ື⏬㓄ಙ䝃䞊䝡䝇
฼⏝⪅ᩘ(YouTube)
3840 x 2160 (Sharp 32inch)
4K
Cinema
ᫎീ㓄ಙ䝃䞊䝡䝇
䛾ᐇ⌧䜈
Raw data: 72Gbps
ᵏᵌᵑಸ
ᵏᵌᵐಸ
26.8ї33.6%
2360ї2900୓ே
Introduction of 3D tech.
Ultra HDTV : 7680 x 4320 (NHK, 145inch PDP)
4
ฟ඾䠖⥲ົ┬㻌 ㏻ಙ฼⏝ືྥㄪᰝ
5
ḟୡ௦䛾䝛䝑䝖䝽䞊䜽䝃䞊䝡䝇䜈
ᑗ᮶䛾䝖䝷䝣䜱䝑䜽
Ultra-High Definition TV 72Gbps(㠀ᅽ⦰᫬)
ᫎീ୺య䛾䝃䞊䝡䝇
-ᫎീᢏ⾡䛾㐍ᒎ
ex. UHD-TV, 3D-TV, 8k/4k Digital cinema
9Layer one VPN㻌 䠄ᑓ⏝⥺䝃䞊䝡䝇䠅
8k/4k Digital cinema
ex. Just-in-time, Optical mesh network
䝛䝑䝖䝽䞊䜽䜈䛾せㄳ
9኱ᐜ㔞䜰䝥䝸䜿䞊䝅䝵䞁
㧗ಙ㢗䞉㉸኱ᐜ㔞
䝝䞊䝗䜴䜵䜰つᶍ๐ῶ
2010’s 2020’s
1990’s
web
2000’s
ᖺ⋡䠖+30-40% = 15ᖺ䛷ᩘⓒಸ
ග䝣䜯䜲䝞㏻ಙ=㉸኱ᐜ㔞
9䝸䜰䝹䝍䜲䝮/䜸䞁䝕䝬䞁䝗
䜰䝆䝸䝔䜱䛾ᐇ⌧
7
ග䝣䜯䜲䝞㏻ಙ䛸↓⥺㏻ಙ䛾ẚ㍑
ග䝣䜯䜲䝞㏻ಙ䛻౑䜟䜜䜛Ἴ㛗䠋࿘Ἴᩘ㡿ᇦ
䢵䢲䢲䢲䢢
䢳䢲䢲䢲䢢
䢸䢲䢲䢢 䢷䢲䢲䢢
䢶䢲䢲䢢
䢵䢲䢲䢢
䢴䢲䢲䢢
࿘Ἴᩘ䣝䣖䣊䣼䣟䢢䢢
ಙྕ䛜✵㛫䛻Ⓨᩓ䛧䛶䛧䜎䛖䛯䜑䚸
ຠ⋡䛜ᝏ䛟䚸➃ᮎ㛫䛾ᖸ΅䛜Ⓨ⏕
ග䝣䜯䜲䝞ෆ䛻ಙྕ䛜㛢䛨䛣䜑䜙䜜䛶
ఏ㏦䛥䜜䜛䛾䛷ຠ⋡䛜㧗䛔
䝣䜯䜲䝞䜢ቑ䜔䛩䛣䛸䛷㏻ಙ㊰䜢ከᩘ
☜ಖ䛩䜛䛣䛸䜒ᐜ᫆
ྍどග䢢
凚䢲䢰䢵䢺到䢲䢰䢹䢹䪱䣯凛䢢
⣸䢢凚⸛凛䢢 㟷䢢
⥳䢢
㯤䢢
ᶳ䢢
㉥䢢
儹儅儈儴僔ᦆኻ䣝䣦䣄䢱䣭䣯䣟䢢
䢲䢰䢳䢢 䢲䢰䢴䢢 䢲䢰䢵䢢 䢲䢰䢶䢢 䢲䢰䢷䢢 䢲䢰䢸䢢 䢲䢰䢹䢢 䢲䢰䢺䢢 䢲䢰䢻䢢 䢳䢰䢲䢢 䢳䢰䢳䢢 䢳䢰䢴䢢 䢳䢰䢵䢢 䢳䢰䢶䢢 䢳䢰䢷䢢 䢳䢰䢸䢢 䢳䢰䢹䢢 䢳䢰䢺䢢 䢳䢰䢻䢢 䢴䢰䢲䢢 Ἴ㛗䣝䪱䣯䣟䢢䢢
䣙䣆䣏㏻ಙ䢢
凚䢳䢰䢶䢴到䢳䢰䢸䢴䪱䣯凛䢢
䢳䢢
䣕䢭䢢 䣕䢢 䣅䢢 䣎䢢 䣎䢭䢢
䢢 䢢 䢢 䢢 䢢
䢢 䢢 䢢 䢢 䢢
䢲䢢
䢲䢰䢵䢺䢢
䢲䢰䢶䢵䢢
䢲䢰䢶䢻䢢
䢲䢰䢷䢷䢢
䢲䢰䢷䢻䢢
䢲䢰䢸䢶䢢
䢲䢰䢹䢹䢢
䢳䢰䢲䢢
䢳䢰䢳䢢
䢳䢰䢴䢢
䢳䢰䢵䢢
Ἴ㛗䣝䪱䣯䣟䢢
䢳䢰䢶䢢
䢳䢰䢷䢢
䢳䢰䢸䢢
䢳䢰䢹䢢
Ἴ㛗䣝䪱䣯䣟䢢
(䝅䝱䝜䞁)䛂㏻ಙᐜ㔞䛿䚸฼⏝ྍ⬟䛺ᖏᇦᖜ䛻ẚ౛䛩䜛䛃
& ග䝣䜯䜲䝞䛿ᴟ䜑䛶ᗈᖏᇦ
10
9
⌧ᅾ䛾㐩ᡂᗘ
z Ἴ㛗ከ㔜䛻䜘䜛኱ᐜ㔞໬䛜ඛ⾜䛧䛶䛔䛯䛜䚸↓⥺㏻ಙ䛷䛿ᙜ↛䛾䛣䛸䛸䛺䛳䛶䛔䜛
䝕䝆䝍䝹䝁䝠䞊䝺䞁䝖 (DSP䛻䜘䜛ఏ㏦㊰➼౯䜔ከ್ุᐃ)ᢏ⾡䛜ᑟධ䛥䜜䛯䛣䛸䛷
100Gbps/Ἴ㛗䛜ၟ⏝䝺䝧䝹䛷ᐇ⌧䛥䜜䛯
z 䛩䛷䛻ᐇ㦂ᐊ䝺䝧䝹䛷䛿䚸100+Tbps/䝣䜯䜲䝞䛜ᐇ⌧ 䛥䜜䛶䛔䜛
‡ D. Qian, “101.7-Tb/s (370㽢294-Gb/s) PDM-128QAM-OFDM Transmission over
3㽢55-km SSMF using Pilot-based Phase Noise Mitigation, ” OFC/NFOEC, PDPB5,
Mar. 2011.
‡ J. Sakaguchi et.al., “19-core fiber transmission of 19x100x172-Gb/s SDM-WDMPDM-QPSK signals at 305Tb/s,” OFC/NFOEC,PDP5C.1, Mar. 2012.
z ୍᪉䛷ග䝣䜯䜲䝞䛾㠀⥺ᙧᛶ䜔ᢞධ䛷䛝䜛ග䝟䝽䞊䛜㝈䜙䜜䜛䛣䛸䛛䜙䚸ග䝣䜯䜲䝞
䛾ᐜ㔞䛻䛿㝈⏺䛜ぢ䛘䛶䛝䛶䛔䜛 (“Capacity Crunch”, 䛂㠀⥺ᙧ䝅䝱䝜䞁㝈⏺䛃)
䝣䜷䝖䝙䝑䜽䝛䝑䝖䝽䞊䜽
z 䝬䝹䝏䝁䜰䞉䝬䝹䝏䝰䞊䝗䛾᳨ウ䛾௚䚸ග䝣䜯䜲䝞䛾ᐜ㔞䜢౑䛔ษ䜛Ⅽ䛾OFDM /
Nyquist WDM 䛾ᑟධ䜔䚸㏻ಙ㟂せ䛻ᛂ䛨䛯䝎䜲䝘䝭䝑䜽䛺Ἴ㛗㈨※䛾๭䜚ᙜ䛶䛜◊
✲䛥䜜䛶䛔䜛䚹
11
⌧ᅾ䛾䝛䝑䝖䝽䞊䜽䛾䜲䝯䞊䝆
Ver.2013.06.06
10
䝜䞊䝗
㛵すᣑ኱ᅗ
180
210
260
280
250 230
270
290 240
300
䝸䞁䜽
20
50
150 60
170 160 200 70
310 180
100 90
260 210 190 110
350 340330
80
280
400
131
370
270 250230 132 120
410
380 360 300290240 220 140
420 430 440 390
320
460
470
450
㏻ಙᣐⅬ(䝜䞊䝗)䛜䝸䞁䜾≧䜒䛧䛟䛿
⥙┠≧䛻ᙇ䜙䜜䛯䝸䞁䜽䛷᥋⥆䛥䜜䛶䛔䜛䚹
30
40
ᡫ̮ܾ᣽ỉỶἳὊἊᴾᵆἼὅἁᵜᵜἠὊἛᵇᴾ
㛵ᮾᣑ኱ᅗ
200
100
190
220
90
110
80
131
132
140 120
JPN48
13
ᖜ䛜ᗈ䛔㧗㏿㐨㊰䛜䚸ᑠ䛥䛺஺ᕪⅬ䛷
᥋⥆䛥䜜䛶䛔䜛䜘䛖䛺≧ἣ
(䝪䝖䝹䝛䝑䜽䠙䝜䞊䝗)
15
䝣䜷䝖䝙䝑䜽䝛䝑䝖䝽䞊䜽
㟁Ẽฎ⌮䛾㝈⏺䜢㉸䛘䛶 䡚Ἴ㛗䝹䞊䝔䜱䞁䜾䡚
㹇㹎࣮ࣝࢱ
Ἴ㛗ከ㔜ࣜࣥࢡ
㸻㌴⥺
㹇㹎ࣃࢣࢵࢺ
䝟䜿䝑䝖䝺䝧䝹䛷䛾㟁Ẽ
ฎ⌮䝹䞊䝔䜱䞁䜾㻌
䊻䞉㐜ᘏ኱䛝䛔㻌
㻌 䞉㐜ᘏ䜖䜙䛞䛒䜚㻌
㻌 䞉኱ᐜ㔞ᅔ㞴㻌
Ἴ㛗ࣃࢫ
㟁Ẽ䝹䞊䝍
ྛ䝜䞊䝗䛷䛾㟁Ẽ䞉ගኚ᥮䜢ᴟຊ᤼䛧䛶䚸䝪䝖䝹䝛䝑䜽䛸䛺䜛㟁Ẽฎ⌮䜢๐ῶ
z Ἴ㛗䛭䛾䜒䛾䜢䝷䝧䝹䛸䛧䛶ಙྕ䜢㆑ู䛧䚸䝹䞊䝔䜱䞁䜾
z ప㐜ᘏ
z ㉸኱ᐜ㔞 (䜲䞁䝍䞊䝣䜵䞊䝇㧗㏿໬䛻䜘䜚䝇䜿䞊䝹䛩䜛)
z ㉸పᾘ㈝㟁ຊ
ග䛾䜎䜎䛾䝹䞊䝔䜱䞁䜾㻌
䊻䞉ప㐜ᘏ㻌
㻌 䞉㐜ᘏ䜖䜙䛞䛺䛧㻌
㻌 䞉኱ᐜ㔞䛾ᐇ⌧㻌
Ἴ㛗ከ㔜ࣜࣥࢡ
㏻㐣䝖䝷䝣䜱䝑䜽䜢㻌
ග䛾䜎䜎⤒㊰ไᚚ㻌
O/E
O/E
O/E
O/E
㹇㹎࣮ࣝࢱ
E/O
E/O
E/O
E/O
LSR
WXC
LSR: Label Switch Router
ගಙྕ䛾┤᥋ฎ⌮䛾㞴䛧䛥
z ග䝇䜲䝑䝏䛾㧗䝁䝇䝖
z 㧗ḟ䝇䜲䝑䝏䛾〇㐀䛾㞴䛧䛥
z ග䝞䝑䝣䜯䛿㝈ᐃⓗ (䝟䜿䝑䝖໬䛻䜘䜛⤫ィከ㔜ຠᯝ䛜ᚓ䛻䛟䛔)
Ἴ㛗࣮ࣝࢸ࢕ࣥࢢ
㻵㻼䝖䝷䝠䝑䜽ኚື䛸㐃
ື䛧䛯䝎䜲䝘䝭䝑䜽䛺
䝺䞊䞁䛾ቑタ䞉๐㝖㻌
WXC: Wavelength Cross-Connect
“HIKARI”䝹䞊䝍
16
17
䝖䝷䝣䜱䝑䜽ቑ䛻䜘䜛䝁䝇䝖ቑ኱
Node cost
Link cost
OXC
OXC
/ROADM
/ROADM
OXC
OXC
/ROADM
/ROADM
OXC
OXC
/ROADM
/ROADM
Traffic increase
Traffic increase
Traffic increase
1. Link cost ๐ῶ (䝣䜯䜲䝞䛾࿘Ἴᩘ㈨※䛾᭱኱฼⏝)
䜶䝷䝇䝔䜱䝑䜽ග䝟䝇䝛䝑䝖䝽䞊䜽
= OFDM䛾ᑟධ䞉ᐦ䛺࿘Ἴᩘ䜾䝸䝑䝗䛾ᑟධ
䝣䜷䝖䝙䝑䜽䝛䝑䝖䝽䞊䜽䛾኱ᐜ㔞໬
2. Node cost ๐ῶ (ග䝇䜲䝑䝏つᶍ䜢๐ῶ)
㝵ᒙ໬ග䝟䝇䝛䝑䝖䝽䞊䜽
=䝟䝇䜢ㄽ⌮ⓗ䛻᮰䛽䜛䛣䛸䛻䜘䜛䝇䜲䝑䝏つᶍ๐ῶ
᪂䛯䛺䝹䞊䝔䜱䞁䜾ᡭἲ䛻ᇶ䛵䛟䝁䞁䝟䜽䝖䛺䝜䞊䝗
Conventional
ITU-T Grid
50GHz
10Gbps
40Gbps
frequency
9 Elastic channel spacing
9 Adaptive modulation
Elastic
Optical Path
Network
frequency
12.5 GHz
ͤ 400Gbps, 1Tbps䜒ど㔝䛻
Designate a set of frequency slots
10
# of frequency slots
Residual bandwidth wasted
100Gbps
Bandwidth
㊥㞳䛻ᛂ䛨䛯ኚㄪ᪉ᘧ䛾㑅ᢥ
䜶䝷䝇䝔䜱䝑䜽ග䝟䝇䝛䝑䝖䝽䞊䜽
40Gbps
19
1 slot : 12.5GHz
8
By introduction of frequency slot and OFDM
6
QPSK
44
By introduction of distance adaptive Modulation
QPSK
QPSK
16QAM
2
ITU-T
nonDA-SLICE
DA-SLICE
0
1
2
3
4
5
6
7 8 9 10 11 12 13 14 15 16
# of hops
Transmission distance
22
Ἴ㛗(⤒㊰)๭䜚ᙜ䛶ၥ㢟
⤒㊰䞉࿘Ἴᩘ䝇䝻䝑䝖๭ᙜၥ㢟
• ᚑ᮶ᆺ䝛䝑䝖䝽䞊䜽䛻䛚䛡䜛“⤒㊰䞉Ἴ㛗๭
ᙜၥ㢟(RWA)” 䛿“⤒㊰䞉࿘Ἴᩘ䝇䝻䝑䝖๭
ᙜၥ㢟(RSA)”䜈
B
A
No.2
D
• Ἴ㛗䜔࿘Ἴᩘ䝇䝻䝑䝖䛾๭䜚ᙜ䛶䛿䚸䜾䝷䝣
⌮ㄽ䛻䛚䛡䜛ᙬⰍၥ㢟䛻┦ᙜ
• ౑⏝࿘Ἴᩘ䛾᩿∦໬䛜኱䛝䛺ၥ㢟䛻
⤒㊰ᅛᐃ䛾ሙྜ
No.1
C
No.1
No.3
No.3
No.2
㞄᥋㡿ᇦ
㐪䛖Ⰽ䛷ሬ䜛
䠎Ⰽ ᚲせ
Fiber
୍᪉䚸⤒㊰๭ᙜၥ㢟䛿ẚ㍑
ⓗᐜ᫆䛻ゎỴྍ⬟䛷䛒䜛
⣽䛛䛺༢఩䛷䛾๭䜚ᙜ䛶䞉୙ᆒ୍䛺౑⏝ᖏᇦ
Wavelength collision
㟼ⓗタィ䞉ືⓗไᚚ䛾཮᪉䛷ᅔ㞴䛥䜢⏕䜐
23
24
Ἴ㛗(⤒㊰)๭䜚ᙜ䛶ၥ㢟
No.1
No.2
C
D
• ㊥㞳䛻ᛂ䛨䛯ኚㄪ᪉ᘧ䜢᥇⏝䛩䜛䜶䝷䝇䝔䜱䝑䜽ග
䝟䝇䝛䝑䝖䝽䞊䜽䛾タィἲ
⤒㊰ᅛᐃ䛾ሙྜ
(i.e. Ἴ㛗๭ᙜ䛾䜏)
B
A
Iterative ILP based Route Assignment [T.Takagi et.al., ECOC2010]
• ୚䛘䜙䜜䛯ග䝟䝇タ❧㟂せ䛾ྛ䚻䛻䛴䛔䛶䚸 ᭱▷
䛾⤒㊰䛷ᚲせ䛸䛺䜛࿘Ἴᩘ䝇䝻䝑䝖ᩘ䜢ィ⟬
No.1
No.3
No.3
No.2
㞄᥋㡿ᇦ
㐪䛖Ⰽ䛷ሬ䜛
䠎Ⰽ ᚲせ
• ୖグ䛾࿘Ἴᩘ䝇䝻䝑䝖ᩘ䛾ぢ✚䜒䜚್䛜ྠ୍䛻䛺䜛
ග䝟䝇㞟ྜ䜢䚸䝇䝻䝑䝖ᩘ䛜ᑠ䛥䛔㡰䛛䜙㡰ḟ⤒㊰
๭䜚ᙜ䛶 => ࿘Ἴᩘ๭䜚ᙜ䛶
Ἴ㛗๭ᙜ䛿䛹䛾⛬ᗘᅔ㞴䠛
z Finding a coloring that minimizes the number of colors
used. => NP-hard
z Finding a coloring that minimizes the number of fibers
used subject to given color set. => NP-complete
25
29
Ratio of Accommodated Traffic
Link cost reduction by path elasticity (static)
The degradation due to the non-uniform path
capacity is marginal.
# of slots necessary for shortest route
ITU-T grid method
ITU-T grid spacing
nonDA-SLICE method
1000
DA-SLICE method
50%
800
nonDA-SLICE
: elastic optical path networks
without distance adaptive modulation
18%
600
400
z
z
z
z
200
6x6 mesh network
Bandwidth of one fiber : 4, 400 GHz
ITU-T grid bandwidth : 100 GHz
Frequency slot : 12.5 GHz
0
1
2
3
4
5
6
7
8
9
Average number of connection demands
10
30
Accepted traffic demand (normalized)
1200
Number of fibers
[T.Takagi et.al., OFC2011]
7x7 polygrid topology
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
ITU-T
nonDA-SLICE
DA-SLICE
Accepted traffic demand (normalized)
Without
fragmentation
1400
[T.Takagi et.al., ECOC2010]
COST266 topology
2.8
2.6
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
ITU-T
nonDA-SLICE
DA-SLICE
Because of the increase in # of wavelength candidates
32
䝣䜯䜲䝞ᐜ㔞ᣑ኱䛾㝈⏺
ග䝣䜯䜲䝞୰
䛾Ἴ㛗䝟䝇ᩘ
Available bandwidth (ex. C+L)
100
100G x 100 = 10T
Nonlinear Shannon limit
Nonlinear impairments
80
40G x 80
400G x 60 =24T
= 3.2T
Transparent optical reach?
Transponder cost?
1
㉸኱ᐜ㔞䝣䜷䝖䝙䝑䜽䝜䞊䝗
䝟䝇ᐜ㔞
10G
40G
100G
400G
䝛䝑䝖䝽䞊䜽඲య䛾᭱㐺໬䛻䜘
䜛䚸ᐇ㉁ⓗ䛺ᐜ㔞ᣑ኱➼䛜ᚲ㡲
Traffic: +30%/year
33
୍㝵ᒙග䝟䝇䝛䝑䝖䝽䞊䜽
Ἴ㛗㑅ᢥ䝇䜲䝑䝏
(Wavelength Selective Switch)
኱ᐜ㔞໬䞉పᾘ㈝㟁ຊ໬ :
1st stage :୍㝵ᒙග䝟䝇䝛䝑䝖䝽䞊䜽
WSS䛾䜏䛷ᵓᡂ䛥䜜䛯
1㝵ᒙ䝜䞊䝗
add
Bottleneck and prevent the constructing costeffective networks
WSS䛾ᵓ㐀
ᅇᢡ᱁Ꮚ
ධຊ䝣䜯䜲䝞
ฟຊ䝣䜯䜲䝞
drop
…
…
…
…
ᑠᆺ䝭䝷䞊
O/E
O/E
O/E
O/E
LSR
: WSS
E/O
E/O
E/O
E/O
LSR: Label Switch Router
(LCOS➼䜒฼⏝)
WXC
WXC: Wavelength Cross-Connect
⌧ᅾ 1x20䜎䛷䛜ၟ⏝໬䛥䜜䛶䛔䜛䛜䚸
䜘䜚㧗ḟ䛺WSS䜢〇㐀䛩䜛䛾䛿ᐜ᫆䛷
䛿䛺䛔
37
୍㝵ᒙග䝟䝇䝛䝑䝖䝽䞊䜽
z 䝛䝑䝖䝽䞊䜽ෆ䛾ග䝟䝇䛿ቑຍ
z ග䝣䜯䜲䝞ᩘ䛿䛭䜜䛻ᛂ䛨䛶ቑຍ
&
z 䝛䝑䝖䝽䞊䜽ᙧ≧䛿ᅛᐃ
9 ᴟపᾘ㈝㟁ຊ
WXC /
ROADM
9 ㉸኱ᐜ㔞
9 ᭱ᑠ㐜ᘏ㻌 ௚
ྠ୍⤒㊰ୖ䜢⛣ື䛩䜛ග䝟䝇䛜ቑ䛘䜛
WXC: Wavelength Cross-Connect
optical fiber
WXC /
ROADM
38
䝖䝷䝣䜱䝑䜽ቑ䛻ᛂ䛨䛶㉳䛣䜛䛣䛸
䝜䞊䝗ᣑᙇ䛾㝈⏺
optical fiber
Finisar♫〇 1x9 WSS (140mm x 220mm)
A possible solution
The issues
㐣ཤ䛾ゎỴ⟇䠖㧗ḟ䝟䝇䜢ᑟධ䛩䜛
Ex.) VC-3/4 in SDH/SONET
9 ග䝇䜲䝑䝏䝫䞊䝖ᩘ䛜⭾኱䛻
9 ᕧ኱䛺ග䝇䜲䝑䝏䜢ᵓᡂ䛩䜛䛾䛿ᴟ䜑
䛶ᅔ㞴
VC-3/4
39
VC-1/2
䝹䞊䝔䜱䞁䜾䛾ຠ⋡ྥୖ
䜽䝻䝇䝁䝛䜽䝖䛾ຠ⋡ྥୖ
䝸䞁䜽ୖ䛾ග䝣䜯
䜲䝞ᩘ䛜ቑຍ
㝵ᒙ໬ග䝟䝇䝛䝑䝖䝽䞊䜽
ᑠつᶍග䝇䜲䝑䝏䛻䜘䜛኱ᐜ㔞䝜䞊䝗ᵓ⠏
」ᩘἼ㛗䛾䜾䝹䞊䝥໬ (㧗ḟ䝟䝇䛾ᑟධ)
䝛䝑䝖䝽䞊䜽䛾ศ๭
୍㒊䛾Ἴ㛗䝟䝇䛾䜏䛜WXC䜢฼⏝
Subsystem modular OXC Node
Architecture (2011ᖺ~)
㝵ᒙ໬ග䝟䝇䝛䝑䝖䝽䞊䜽 (2000ᖺ㡭~)
z 」ᩘ䛾Ἴ㛗䜢ㄽ⌮ⓗ䛻䜾䝹䞊䝥໬(䛂Ἴ㛗⩌䝟䝇䛃)
z ྍ⬟䛺㝈䜚䜾䝹䞊䝥༢఩䛷䝹䞊䝔䜱䞁䜾
z Ἴ㛗⩌䝟䝇䜢஌䜚᥮䛘䜛䛸䛝䞉add/drop䛾᫬䛻୍㝵ᒙ
ᆺ䛸ྠᵝ䛾ฎ⌮䜢ᐇ᪋
z ኱つᶍ໬䛻㝈⏺
optical fiber
z ྛ䝛䝑䝖䝽䞊䜽䝜䞊䝗䜢䚸ᑠᆺ
OXC䝃䝤䝅䝇䝔䝮䜢୪䜉䛯䜒
䛾䛷ᵓᡂ
z ྛ䝃䝤䝅䝇䝔䝮䛿ᐜ᫆䛻ᵓ
ᡂྍ⬟
WXC
waveband path
BXC
wavelength path
Grouped Routing Entity based Optical Networks (2011ᖺ~)
z Ἴ㛗⩌༢఩䛷䛾⤒㊰ไᚚ䛸Ἴ㛗༢఩䛾add/drop䜢
⤌䜏ྜ䜟䛫䛶䛔䜛
z Ἴ㛗⏝䜽䝻䝇䝁䝛䜽䝖䛿୙せ䛷䝜䞊䝗䛜䝁䞁䝟䜽䝖
BXC: WaveBand Cross-Connect
Waveband Path = a group of wavelength paths
z ྍ⬟䛺㝈䜚Ἴ㛗⩌䝟䝇༢఩䛷䝹䞊䝔䜱䞁䜾
2-stage Routing Optical Networks (2011ᖺ~)
z ḟ䛾⤒⏤䝜䞊䝗䛻ᛂ䛨䛶䚸Ἴ㛗䜢ືⓗ䛻䜾䝹䞊䝥໬
z ⥆䛟ᑠᆺ䛾ග䝇䜲䝑䝏䛷䚸ග䝣䜯䜲䝞䜢㑅ᢥ䛩䜛
z WXC䛿౫↛䛸䛧䛶኱䛝䛟䚸኱つᶍ໬䛻㝈⏺
41
Hierarchical Optical Path Network
Benefit of Wavebands
Waveband Path = a group of wavelength paths
Optical Fiber
Waveband Path
42
Ratio of the total number of
switch ports in the networks
(R: Hierarchical/Single Layer)
WXC
Wavelength Path
bandwidth, W
2.5
waveband hops, H
2
2.5
BXC
• Reduce port counts
• Large capacity optical paths
1.5
BXC: WaveBand Cross-Connect
1
wavelength path
Single-layer
2
1.5
1
0.5
0.5
0
WXC
WXC
WXC
WXC
WXC
WXC
0
4
Hierarchical
waveband path
BXC
BXC
Number of optical
switch ports decreases
over a wide area
8
12
16 1
BXC
2
3
4
5
6
7
8
43
Normalized network cost
䠄Hierarchical / Single-layer network䠅
Performance Evaluation [I.Yagyu et.al. 2008]
1.0
㝵ᒙ໬ග䝟䝇䜽䝻䝇䝁䝛䜽䝖䝜䞊䝗䛾ヨస
㛗㇂ᕝᾈ㻌 బ⸨೺୍ (ྡྂᒇ኱Ꮫ)㻌 㧗ᶫᾈ(NTT䝣䜷䝖䝙䜽䝇◊✲ᡤ)㻌 ዟ㔝ᑗஅ(NTT䜶䝺䜽䝖䝻䝙䜽䝇)
㟁Ꮚ᝟ሗ㏻ಙᏛ఍䝋䝃䜲䜶䝔䜱኱఍2010 ᣍᚅㅮ₇
2.0
1.5
44
end-to-end
BPHT (ͤ)
Proposed
Single-layer
0.5
0
0
1
2
3
4
5
6
7
8
Average number of wavelength paths between node pairs
ͤ) BPHT : X.Cao et al., IEEE J-SAC ,2003
OFC2010 Post-deadline Paper
4/6䝥䝺䝇Ⓨ⾲㻌 䠄᪥⤒⏘ᴗ᪂⪺, ᪥หᕤᴗ᪂⪺, 䝣䝆䝃䞁䜿䜲䝡䝆䝛䝇䜰䜲, ୰᪥᪂⪺, ⛉Ꮫ᪂⪺䛻ᥖ㍕䠅
㝵ᒙ໬ග䝟䝇䝛䝑䝖䝽䞊䜽
Grouped Routing Scheme
Dropped
Wavelength path
୍㒊䛾Ἴ㛗䝟䝇䛾䜏䛜WXC䜢฼⏝
Added
Wavelength path
䝹䞊䝔䜱䞁䜾䛿␯䛺⢏ᗘ䛷ᐇ᪋
(Ἴ㛗⩌䝟䝇䛸ྠᵝ)
¾ 䝣䜯䜲䝞୰䛾Ἴ㛗䝟䝇䛿䚸ᗄ䛴䛛䛾䜾䝹䞊
䝥䛻ㄽ⌮ⓗ䛻ศ๭ (GRE).
¾ BXC䛿GRE༢఩䛷䛾䝹䞊䝔䜱䞁䜾䜢ᢸᙜ
Grouped Routing Entity (GRE)
BXC
wavelength path
…
…
WXC
waveband path
…
optical fiber
…
GR-OXC
= The bundle of wavelength paths used
for coarse granular routing
GRE
BXC: WaveBand Cross-Connect
Waveband Path = a group of wavelength paths
1x2 WSS
1x2 WSS or SC
Coupler
1xN WBSS
z ྍ⬟䛺㝈䜚Ἴ㛗⩌䝟䝇༢఩䛷䝹䞊䝔䜱䞁䜾
Grouped Routing
z WXC䛿౫↛䛸䛧䛶኱䛝䛟䚸኱つᶍ໬䛻㝈⏺
= Coarse granular routing with
Fine granular add/drop
Add/drop᧯స䛿⣽⢏ᗘ䛷ᐇ᪋
(Ἴ㛗༢఩)
¾ ௵ព䛾Ἴ㛗䜢GRE䜘䜚ᢳฟ䞉ᤄධྍ⬟
¾ ᢳฟ䞉ᤄධ᫬䛻䛿䚸GRE䛜⤊➃䛩䜛ᚲせ
䛿↓䛔 (“GRE䝟䜲䝥”)
47
48
Grouped Routing Scheme
Destination
GRE pipe
9 No termination functions are defined
9 GRE pipe is NOT “path” of ITU-T definition
Source
GRE pipe
fiber
GR-OXC
኱ᐜ㔞ග䝹䞊䝔䜱䞁䜾⏝䝕䝞䜲䝇
Wavelength path
¾ Virtual Any
pipes
carrying
multiple wavelength
paths
connectpaths
multiple nodes.
node
can accommodate
multiple
wavelength
¾ Wavelength
pathsdifferent
can be added/dropped
to/from
virtual
pipes at any
having
s-d node pairs into
a GRE
pipe.
arbitrary node.
51
Ἴ㛗㑅ᢥ䝇䜲䝑䝏
Req. on WBSS’s spec [K.Takaha et.al. 2013]
‹ Higher degree
¾ Pan-European network (COST266)
Ἴ㛗㑅ᢥ䝇䜲䝑䝏
Wavelength Selective
Switch, WSS
Min
26
Min
2
Link distance Max 1712
Node degree Max
8
Ave. 627.2
Ave. 3.92
Number of links
z ධຊ䝫䞊䝖䛛䜙䛾Ἴ㛗ಙྕ䜢䚸Ἴ㛗ẖ䛻ᡤᮃ䛾ฟຊ䝫䞊䝖
䜈䝹䞊䝔䜱䞁䜾䛩䜛
51
Hop count
Max degree
200
Number of nodes
Min
1
Max
6
Ave. 2.76
Brussels
COST266 network
M : wavelengths of GRE capacity
‹ More wavelength paths
¾ 88 + 8 wavelengths 50GHz spacing
( C-band + a part of S-band, L-band )
‹ Broader WB/GRE capacity
‹ Compact implementation
z 䝟䝽䞊ㄪᩚ➼䚸ᵝ䚻䛺ᶵ⬟䜢୍䛴䛾⟽䛻ワ䜑㎸䜣䛷䛔䜛
z ఱྎ䛛䛾WSS䜢᥋⥆䛩䜛䛣䛸䛷䚸ග䜽䝻䝇䝁䝛䜽䝖䛜ᵓᡂྍ⬟
z 2005ᖺ㡭䛛䜙Ⓩሙ䚸⌧ᅾᛴ⃭䛻ฟⲴ㔞䜢ఙ䜀䛧䛶䛔䜛
52
Ratio of requred
path # of selective switch
†
Y.Taniguchi et al., JOCN 2013
†
54
Proposed 8x8 WBXC
Proposed 1x8 WBSS
Four WBSSs module
8x8 WBXC
WBSS
WB1
switch
WB2
switch
WB3
switch
WB4
switch
z
WB5
switch
WB6
switch
WB7
switch
WB8
switch
1x8 cyclic AWG
チップサイズ
74.6mm x 48.4mm
チップサイズ
74.6mm x 48.4mm
Transmission characteristics
1x8 Optical switch
Property
center wavelength error
insertion loss
channel loss deviation
in each output loss
polarization dependent loss
1dB channel bandwith
3dB channel bandwith
adjacent crosstalk
non‐adjacent crosstalk
coherent crosstalk
30 mm
Value
‐0.04 to 0.02 (nm)
4.47 to 7.69 (dB)
Specifications
1.56 to 2.10 (dB)
0.02 to 0.41 (dB)
> 0.10 (nm)
> 0.18 (nm)
< ‐ 37.03 (dB)
≤ ‐ 37.67 (dB)
≤ ‐ 32.19 (dB)
 Adaptation to 50 GHz spacing signals
 Larger WBSS Scale (1x5 WBSS → 1x8 WBSS)
by very small increase of WBSS PLC chip’s size
1x8 WBSS chip
Optical coupler
55
WBSS
 8x8 WBXC
 96 wavelengths / fiber
•50 GHz spacing on ITU-T grid
• 8 wavebands / fiber
•12 wavelengths / waveband
 Throughput 7.68 Tbps
 Compared to previous device (5x5 WBXC)
 Number of ports : 1.60 times
 Size of WBXC module : 0.68 times
 Capacity of wavelengths : 1.50 times
56
Experiment:spectra
まとめ
57
58
結論
 通信量は依然として増加している(+30‐40%/年)。
 エネルギー消費・装置コストには限度がある。
 超低消費電力かつ超大容量を実現する上では、フォト
ニックネットワークを導入していくことが必要である。
 フォトニックネットワークの導入では、依然として高価な装
置コストをどのように抑制するかが鍵である。
ご清聴ありがとうございました!!
 限定的な能力しか持たない装置・ネットワークと、その特
性を考慮した最適化手法を組み合わせることで大きな性
能アップを実現できる。
 ネットワークアーキテクチャ・新たな光デバイスの開発に
より、大幅な容量拡大が可能となった。
59
謝辞:本研究の一部は KAKENHI (23246072)
およびSCOPE により実施された。
60
Fly UP