《全光 IP 网：神话还是现实？——Juniper.ppt》由会员分享，可在线阅读，更多相关《全光 IP 网：神话还是现实？——Juniper.ppt（20页珍藏版）》请在三一办公上搜索。
1、,ALL OPTICAL IP NETWORKS,Myth or reality?,Luc Ceuppens,Vice President,Product MarketingInfrastructure Product Group,2,Copyright 2010 Juniper Networks,Inc.,OUTLINEOptical bypass:what,why and how?The forgotten piggyChallengesMyths and realitiesConclusion and summary,3,Copyright 2010 Juniper Networks,I
2、nc.,OPTICAL BYPASS:AN INTRODUCTIONNetwork planning technology where optical connectivity istopologically richer than IP/MPLS node map.IP/MPLSnodes are by-passed by Select wavelengths inside the transport optical element(optical bypass)OTN circuits(opaque bypass,OTN bypass)Often popularized as a tool
3、 to reduce network CAPEX andenergy footprint Reduction of transit traffic(and hence ports)on IP/MPLS node On shorter segments(less than 1,500km)a“bypass”wavelength can beoptically amplified in the node without OEO conversion,further reducingthe costIn essence,the promise is that bypass shifts the CA
4、PEXbudget out of IP/MPLS equipment towards optical gear.,4,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:GOOD IDEA?OB is backed up with significant amount of research data andwhitepapers.Several peer-reviewed CAPEX network modelstudies were done to prove OB monetary/energy savings.List of popu
5、lar titles include“Evolution of WDM Optical IP:Cost and Energy”(R.Tucker)“Packet Layer Topologies of Cost Optimized Transport”(NSN)Deliverables within European“IST Nobel 2”project and othersSeveral service providers(SwissCom,DT,FT)have endorsedoptical bypass in presentations and articles in optical-
6、themedmagazines.OB also enjoys significant analyst coverage.Note 1:OB is sometimes referred to as“optical circuit switching”OCS,or“optical cross-connect”OXC)Note 2:This presentation uses several graphics from above mentioned papers,5,Copyright 2010 Juniper Networks,Inc.,COMPONENTS OF OPTICAL BYPASS1
7、.WDM/ROADM shelves(in red)serving IP/MPLS nodes2.Selective wavelength routing in the optical layer,6,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:ONLY THE BEGINNING?Transition from all-electrical switching(“opaque”)in everynode to OB/OCS(“transparent”)is often viewed as precursorto all-optica
8、l packet processing:In this“Stage 1”,OCS elements are often shown operatingunder G-MPLS control for dynamic path provisioning,7,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:ELEVATOR PITCH1.IP/MPLS port cost(in abstract units)2.OXC/ROADM port cost(in abstract units)Replace PACKET switching by
9、OPTICAL switching=LOWER COST!,8,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:ELEVATOR PITCH(CONTD)Did anyone notice a pink piggy?If not,lets try again,9,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:THE FORGOTTEN PIGGYThere is a“pink piggy”in the network costpicture,so large that people
10、 barely notice it.It is the same piggy that drove thetransformation from circuit to packet as well asIP convergence.This piggy is called“cost of bandwidth”.It feeds upon available bandwidth and returnsmonetary services.It heavily favors statisticalmultiplexing over circuit switching because it canea
11、t much less and monetize much more.So,why do some people ignore it?,10,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:CHALLENGESThe smallest unit in OCS is a“wavelength”.Optical Bypass is contradictory to service providers desire toimprove spectral efficiency.The higher the wavelength speed,the
12、 less opportunityfor bypass Literature actually acknowledges thisTraffic matrices tell a different story than bandwidth matricesDynamic provisioning of a light path thru a network of opticalswitches(e.g.“packet over light LSP”)requires complexpower/amplification computations for OXC nodes and cantak
13、e several minutes,11,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:MYTH 1IP/MPLS NODES IN THE CORE ARE ROUTING PACKETSThe packet route is already known at the core ingress point.However,packet core switches are NOT doing routing theyare intelligent statistical multiplexors filling the core ban
14、dwidthat packet-level resolution.Opportunity to replace core routers by lower cost LSRs,Edge12,Core,Core,Inner CoreCopyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:MYTH 1IP/MPLS NODES IN THE CORE ARE ROUTING PACKETSPacket network evolutionOtherNetworksOtherNetworks,13,Copyright 2010 Juniper Netwo
15、rks,Inc.,OPTICAL BYPASS:MYTH 2TRANSIT TRAFFIC CAN BE EFFICIENTLY“REMOVED”,14,OPTICAL BYPASS:MYTH 2TRANSIT TRAFFIC CAN BE EFFICIENTLY“REMOVED”Traffic in real networks follows complex distribution patternsand static de-multiplexing wastes bandwidthTraffic BC,OverloadReduced peak rate,X/2,X/2,Overload,
16、Overload,Traffic ACBandwidth XOptical split option(b),UnutilizedUnutilized,Unutilized,Unutilized,Unutilized,Unutilized,Unutilized,Unutilized,X,XCopyright 2010 Juniper Networks,Inc.,UnutilizedOptical split option(a),15,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:MYTH 3OPTICAL BYPASS SAVES IP/
17、MPLS PORTSBy off-loading transit traffic,the need for IP/MPLS ports isreducedWDM network operates at wavelength speeds.A router thatneeds to talk to N peers over fully transparent network needsN ports with capacity equal to lambda speed Number of IP/MPLS ports grows!Doctored traffic volumes and low
18、inter-node connectivitydegree are the mandatory conditions to demonstratereduction in requirements for IP/MPLS ports,16,Copyright 2010 Juniper Networks,Inc.,OPTICAL BYPASS:MYTH 4“TRANSPARENT”NETWORK SCALE TO COMPLEX TOPOLOGIESClaim:all-optical cross-connect network can link many packetnodes over ded
19、icated point-to-point light pathsFull-mesh in optical domain requires N*(N-1)/2 lightpaths,where N is the number of nodes.For WDM density of80 channels,networks as small as 10-15 nodes can easilyrun of out of channelsCost would be prohibitive!Complexity of edge devices increases exponentially,17,Fai
20、ledCopyright 2010 Juniper Networks,Inc.,Unusable,U,OPTICAL BYPASS:MYTH 5“TRANSPARENT”NETWORKS OFFER GOOD RESILIENCYClaim:all-optical cross-connect network offers high availabilityIf a light path that crosses several nodes has one brokensection,the entire path becomes unusable.OXC network canonly off
21、er 1:1 protection,effectively doubling provisionedbandwidth.Dynamic control plane(G-MPLS)cannot help because thereare no traffic sources suitable to fill the surviving sectionsUnusable,18,Copyright 2010 Juniper Networks,Inc.,SO,WHAT IS THE VERDICT?“Optical bypass”offers minimal traffic processing at
22、 the cost ofinefficient bandwidth utilization.This is a valid design compromisein the following cases:Optical spectrum is free(dark fibers or lambdas)There are intermediate nodes with very small tributary trafficNetwork designers considering OB should remember the piggy!Cost of bandwidth can not be
23、ignored Statistical multiplexing maximizes bandwidthutilizationInstalling expensive packet nodes is justified by thecost savings in other network resources such as linkcapacity.Our tests show that it pays off to install packet nodes evenif the cost is far greater than the link cost.“Multi-layer MPLS
24、 Network Design:The Impact of Statistical Multiplexing”Dr.Pietro Belotti,Carnegie Mellon University,2008,19,Copyright 2010 Juniper Networks,Inc.,SUMMARYAll-optical networks offer very poor bandwidth utilization andwill remain in Power Point for most network operatorsResponsible use of optical bypass
25、 in limited cases iswelcome in network design,provided the cost of extralambdas and resiliency has been properly accounted for.Beware of network planning models,where the cost ofbandwidth and effects of statistical multiplexing are ignored.The poor economy of“optical bypass”is recognized by mostof OB vendors,which silently“complement”OXC withelectrical-domain switching(OTN).,
