strategy for switch's next generation optical network · strategy for switch's next...

31
Strategy for SWITCH's next generation optical network Terena Network Architecture Workshop 2012 Felix Kugler, Willi Huber [email protected] [email protected] 21.11.2012

Upload: vuhanh

Post on 04-Aug-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

© SWITCH 2012

Strategy for SWITCH's next generation optical network Terena Network Architecture Workshop 2012

Felix Kugler, Willi Huber [email protected] [email protected]

21.11.2012

© SWITCH 2012

Introduction !The project for the SWITCHlan renewal was formally accepted on November 15, 2012.

•  no detailed concept is available yet •  we present our vision, goals, and the

rough concepts

2

© SWITCH 2012

Agenda!

•  where we are today •  where we want to go •  how to get there

– fibers and optical transmission – switching layer

•  summary

3

© SWITCH 2012

•  11 years since first light •  2500 km of fiber pairs •  ~60 PoPs

SWITCHlan mid 2012!

6

•  high quality, high performance, low latency IP backbone •  multiple upstreams to commercial Internet @ 2 large

exchange points •  exclusive GEANT access

“hotspots”: areas with largest bandwidth needs

regional links •  1Gbps •  few with 10Gbps

core backbone •  DWDM •  10..16 lambdas •  max 10Gbps/lambda

© SWITCH 2012

SWITCHlan is doing fine!!•  IP service

–  0.1..20 Gbps access speeds

•  OPN (optical private network) –  EoMPLS based P2P and multipoint links (0.1..10Gbps) –  resilient –  no bandwidth guarantee (but overprovisioning and careful monitoring)

•  lambda services –  unprotected only –  so far exclusively used for LHCOPN CBF links –  sub-standard availability !

7

CNAF - KIT

CNAF - CERN CNAF - FERMILAB

our No.1 service!

© SWITCH 2012

Why fix what ain’t broken ?!

SWITCHlan’s DWDM gear is getting old •  some components face their end of life •  number of free lambdas decreasing rapidely •  investments no longer pay out •  no development to 40G/100G •  inflexible operations

8

…though less than 50% of lambdas are in use today!

© SWITCH 2012

Technology has gone a long way !

9

optical switching (ROADMs)

2005

2012

size

sophisticated modulation schemes

tunable optics

Let us make best use of it!

© SWITCH 2012

Agenda!

•  where we are today •  where we want to go •  how to get there

– fibers and optical transmission – switching layer

•  summary

10

© SWITCH 2012

What we anticipate…!•  new users to be connected, mostly small ones •  continued average traffic growth of 40%, and •  sudden high bandwidth demands triggered by outplacement

of storage & computing •  only few users will need full 100G any time soon •  10G will prevail for the next years •  new trend: sensor sites at far away locations

–  bandwidths needs exceeding GPRS, ADSL range –  but very limited budget –  often not permanent

11

© SWITCH 2012

Long Term Development Strategy

•  optical layer –  forced upgrade all our DWDM infrastructure

within 3 years (amplifiers, ROADMs, filters etc) – a long term investment (for a 10 years life) – additional channels and new functionalities

deployed when needed – select a vendor with

•  a clearly defined, realistic road-map •  willingness for a continuing, close collaboration

13

© SWITCH 2012

Long Term Development Strategy

•  network layer – evolutionary approach – continue to use existing (10G capable) routers – new 100G capable router will be deployed as

needed – shorter renewal cycle compared to the optical

equipment

14

© SWITCH 2012

Four things we want to improve!•  bandwidth reserves

–  more lambdas –  100Gbps capability

•  flexibility –  transparent lambdas between all core nodes –  rapid channel setup –  auto-calibration of power levels

•  stability –  topology improvements to enhance resilience –  optical switching capability at major junctions

•  coverage –  enlarge fiber footprint –  extend core backbone over whole country

15

© SWITCH 2012

Agenda!

•  where we are today •  where we want to go •  how to get there

– fibers and optical transmission – switching layer

•  summary

16

© SWITCH 2012

Overall network stability!•  based on 10 years experience we postulate

–  three independent paths into “hotspots” Zürich, Geneva, Lugano

–  two-way resilience on the core backbone –  single fiber paths sufficient within city limits

•  stability of electric power worse than availability of fibers –  crucial: keep connectivity up even if a whole region goes black

for a long time –  run two powerful exit points to the Internet –  build bypasses around major network hubs

17

© SWITCH 2012

Proposed fiber footprint 2015 !

19

moderate expansion of fiber footprint

© SWITCH 2012

Optical transport 2001..2012 !

20

Sorrento DWDM 2001..2003 16 lambdas

BTI DWDM 2005..2010 10 lambdas

CWDM router plugins + passive optics 1..4 lambdas

© SWITCH 2012

The new optical core!•  equal capabilities for all regions •  uniform transmission gear •  to be built on existing fibers, while services remain up •  connects all cross border links to neighbor NRENs

21

GARR

Renater

DFN BelWü

extend core backbone

© SWITCH 2012

Photonic core backbone design!

•  most nodes add/drop few lambdas only •  express lambdas between network hotspots prevail

–  minimizing OEO conversions seems reasonable –  thus we favor photonic ROADMs

•  no bundling or sub-lambda switching on optical layer 23

typical span 60..100km

large university or computing center

large university or computing center

smaller university

max ring length 1000km

transponders

core nodes

numerous express lambdas between network hotspots network “hotspot”

© SWITCH 2012

Key requirements!•  40 lambdas, easily upgradable to 80 •  native 10G and 100G wavelengths •  1000km reach without regeneration •  auto-calibration of optical power levels •  support for alien waves •  tolerance for optical signals outside C-band •  smooth path toward super-channels for

400G compatibility appreciated

24

© SWITCH 2012

Optical switching !•  optical switching capability mandatory •  initial deployment of multi-degree ROADMs

–  at selected major fiber junctions –  at entry points of native lambda services

•  long-term plans remain to be defined –  use of direction-, color-, contention-less ROADMs – where and when ? –  functionality versus reliability ? –  economics, size, complexity ?

25

© SWITCH 2012

About economics!•  equipment, connectivity

– dramatic increase in performance & functionality per $$

•  maintenance – cost approx. stable

•  fiber leases (CH) – 45% of overall network cost –  last 10 years: ~30% decrease in lease cost/m – won’t go significantly down any more

26

find ways to mitigate!

© SWITCH 2012

Fiber sharing with Hybrid WDM!

27

core backbone •  40..80 DWDM channels •  OSC •  spans up to 100km •  unrepeatered links up to

1000km

1503..1577 [nm] 1470 1490 1590 1610

regional links •  4 CWDM channels •  links up to 60km

L-Band range, suitable for amplified long range links

for use •  by SWITCH to connect small sites •  by regional networks

Use existing backbone fibers to connect small sites!

© SWITCH 2012

Fiber sharing with Hybrid WDM!

•  possible for a few dB more span loss… •  micro-housing at regional junctions: cheap, small, passive

28

DWDM node

DWDM node

core site core site

(routerless)

small site

HW

DM

split

ter

HW

DM

split

ter

HW

DM

split

ter

HW

DM

split

ter

micro-housing

© SWITCH 2012

Agenda!

•  where we are today •  where we want to go •  how to get there

– fibers and optical transmission – switching layer

•  summary

29

© SWITCH 2012

IP Layer Design Goals •  network services

–  99.99% availability –  overprovisioning -> virtually no packet loss –  lowest possible transfer delays –  IP service & private layer 2 networks

•  network infrastructure –  meshed network –  two exit points at the extreme (Geneva & Zurich) –  PoPs at customer premises

•  border of responsibility: Ethernet interface •  footprint & power is an issue

–  must scale to small and big sites (1 : 100)

30

© SWITCH 2012

Routing vs. Switching •  shortest (physical) path first:

–  IP routers –  shortest path bridging (TRILL, IEEE 802.1aq)

•  test with TRILL (Brocade switches): –  missing network management tools –  lacking interoperability among vendors –  difficult to provide layer 2 services –  doesn’t scale for a national network

⇒  routed IP backbone –  robust, proven and well scalable solution –  small organizations connected with stub links (without router)

31

© SWITCH 2012

Layer 1 & 2 Services •  emulated layer 2 services

– benefit from meshed backbone topology – EoMPLS point-to-point links, VPLS multipoint

networks

•  lambdas – only for higher bandwidths, 10G and 100G – protection will be available

32

© SWITCH 2012

MPLS in the Backbone •  layer 2 services forced us to deploy MPLS in

the backbone –  the only available tunneling technology, that was

implemented in hardware –  used for a restricted address range for layer 2

services only

•  desirable solution: –  tunneling technique using encapsulation in IP –  no need for additional protocols in the backbone –  a use case for SDN?

33

© SWITCH 2012

Agenda!

•  where we are today •  where we want to go •  how to get there

– fibers and optical transmission – switching layer

•  summary

34

© SWITCH 2012

Summary •  fiber footprint

–  modest expansion

•  transmission layer –  core backbone to cover all regions –  advanced, uniform DWDM system on core

•  built to last 10 years •  coexists with other optical services on same fibers

•  switching layer –  simple, robust, high performance routed IP network –  overprovisioning, no BoD –  private networks nevertheless needed

•  traffic isolation •  no traffic engineering

35

© SWITCH 2012

Strategy for SWITCH's next generation optical network Terena Network Architecture Workshop 2012

21.11.2012