Zhao Feng

（Institute of Comm unication Standards Research,China Academ y of Telecommunication Research, Ministry of Industry and Inform ation Technology of the Peop le's Republic of China, Beijing 100083, P. R. China）

Abstract:In order to compare the effectiveness of d ifferent Carrier Ethernet（CE）technologies,testmethods are desc ribed in term s of the features of CE defined by Metro Ethernet Forum（MEF）,inc lud ing scalability,reliability,Quality of Service（QoS）,standard ized services,and service managem ent.It can be seen from the test resu lts that the p resent CE technolog ies could satisfy basic requirements of carrier-class network on the abovem entioned aspects.However,comp lete interworking cou ld hard ly be achieved between d ifferent CE solutions in the short term because o f their g reat d ifference in service bearer capab ility,reliab ility,scalability,QoS and Operation,Adm inistration and Maintenance（OAM）ability.

1 Concept of Carrier Ethernet

The Metro Ethernet Forum（MEF）first introduced the concep tCarrier Ethernet（CE）and began the research ofCE.Accord ing to MEF,CE has five features:

（1）Scalability

·Services are scalab le,allow ing m illions ofusers to access the same network service;

·Bandw id ths are scalab le,allow ing the rate to range from 1Mb/s to 10 Gb/s oreven higher.The rate can increase by certain g ranularity.

（2）Carrier-Class Reliability

·Provide a p rotection sw itching time ofnomore than 50ms;

·Be capab le ofend-to-end path p rotection;

·Supportagg regation link p rotection and node p rotection.

（3）Quality ofService（QoS）

·Deliverguaranteed end-to-end performance;

·Enab le selection ofend-to-end QoS c lasses;

·Be app licab le to business,mobile and access convergence scenarios.

（4）Standard ized Services

·Provide Ethernetand VirtualLAN（VLAN）services;

·Seam lessly integ rate Time Division Multip lexing（TDM）services;

·Supportcircuitemulation services;·Supportexisting voice app lications.（5）Service Management·Supportquick service delivery;·Supportcarrier-c lass Operation Adm inistration and Maintenance（OAM）;

·Provide the customernetwork management func tion.

2 CE-Related Tests

CE has become a hot research top ic in the telecom industry.Many international standard ization organizations,such as ITU-T,IEEE,IETFand MEF,have started CE-related standardization.Moreover,there have already been several solutions in themarket.They are rep resented by Provider Backbone Bridge（PBB）/Provider Backbone Transport（PBT）［1］,TransportMultip rotocol LabelSw itching（T-MPLS）［2］,enhanced Ethernet technology and VirtualPrivate LAN Service（VPLS）［3］.The five features ofCE can be used as the criteria to evaluate these technologies and related tests should be conducted.

2.1 Scalability

The ob jective ofscalability test is to verify the bandw id th and tunnelp rocessing capability a CE solution can p rovide.

The bandw id th testmethodology described in RFC2544［4］can be used to verify whethera System Under Test（SUT）supports w ire-speed forward ing through FastEthernet（FE）,Gigabit Ethernet（GE）,and 10GE interfaces.

The tunnelp rocessing capability test ismainly used to verify themaximum numberof tunnels,both innerand outer,thatcan be configured w ith a SUT,as wellas themaximum services the SUT supports in case ofa sing le tunnel（i.e.multi-servicemultip lexing tunnel capability）.The tunnelp rotec tion mechanism should be enab led for the SUTand only the results w ith a p rotection sw itching time ofnomore than 50ms make sense.The identifications of tunnels vary w ith SUTs.Forexam p le,in PBT,the outer tunnelis identified w ith Backbone VLAN（B-VLAN）and the inner tunnelis identified w ith I-SID;in T-MPLS and VPLS,both innerand outer tunnels can be identified w ith MPLS labels;and in enhanced Ethernet technology,inner and outer tunnels are identified w ith inner and outer VLAN labels of Provider Bridge（PB）.

2.2 Reliability

The ob jec tive of reliability testis to verify the tunnelp rotection sw itching time ofa SUT,seeing if itmeets the carrier-c lass requirement,i.e.nomore than 50ms.

The p rotection sw itching time is calculated as follows:

Protection sw itching time=Number of lostpackets/Transm it rate ofpackets

The end-to-end tunnelsw itching may be triggered in any of the follow ing conditions:link failure,port failure,and node failure.These triggering conditions should be verified respectively,and the operations involve disconnecting op tic fiber,turning off the forwarding port,powering off the device and unp lugging the ac tive board.The p rotection sw itching test,emulates the actual network based on the equipment configuration w ith a large quantity of tunnels and heavy data flows thatare evenly distributed to all tunnels.

The network topology used for reliability testcan be ring,g rid or dual-hom ing,depend ing on the technology adop ted.In case of ring networking mode,it is required to test suchmultiring topologies as intersecting and tangent topologies.

The sw itching p rotection testshould be conducted forunicastand multicast tunnels respectively so as to verify the p rotection of the SUT forb roadcast services and multicastservices.

Besides,it is necessary to verify ifany packetis lostduring the tunnelis sw itched back after the failure is recovered.

2.3 QoS

The ob jective ofQoS testis to verify whethera SUT canmeet the requirements for carrier-c lass QoS.QoS testinvolves such aspec ts as access controlpolicy,traffic labeling policy,adm ission controlpolicy,and queue d ispatching mechanism.

The access controlpolicy testaim s to find outwhether the SUT can correctly configure access controlpolicies for L2,L3 and L4 traffic and whether the forwarding performance is affec ted after the Access Control List（ACL）is configured.

The traffic labeling policy test is to see whether the SUT canmark or remark a data flow w ith a QoS c lass according to the p redefined traffic labeling policies.The QoS c lassification can be based on IEEE 802.3,802.1ah,MPLS or IPType of Service（ToS）/Differentiated Services（DiffServ）.

The adm ission controlpolicy testis to check whether the SUT can,based on Comm itted Information Rate（CIR）or Peak Information Rate（PIR）,execute rate controlover the access flows,as wellas to check the bandw id th g ranularity the SUT can control.

The queue d ispatching mechanism test is to testwhether the SUTsupports such p riority queue d ispatching mechanism s as Priority Queuing（PQ）,Weighted FairQueuing（WFQ）,or Weighted Round Robin（WRR）.

In add ition,during QoS test,it is required to verify sing le-layerand layered QoS strategies respectively.

MEF14［5］is reference for the QoS test.

2.4 Service Support Capability

The ob jec tive ofservice support capability test is to verify whether the SUTsupports such service types as Ethernet Line（E-Line）,Ethernet LAN（E-LAN）,E-TREE and TDM.

In the E-LAN service test,itis required to pay specialattention to the method the SUT creates E-LAN services and theway it learns remote Med ia Access Control（MAC）add resses.

The E-TREE service testis used to verify whether the leafnodes can be isolated from each otherand whether multicast function can be achieved via E-TREE service.

In TDM service test,it is required to measure such performance indexes as delay,jitterand error code rate.The test portcan be framed E1,non-framed E1 or channelized Synchronous Transport Module（STM）-1,depend ing on the configured devices.

Besides,it is necessary to verify whether the SUTsupports TDM c lock synchronization,inc lud ing c lock synchronization in CircuitEmulation Service（CES）,c lock synchronization of Ethernet,and time synchronization,specified in IEEE 1588 standard.

MEF9［6］,MEF18［7］and MEF19［8］are reference for the service supportcapability test.

2.5 Service Management

Servicemanagement testmainly involves two aspects:Network Management System（NMS）and OAMmechanism.

The test for NMS should be conducted based on the Fault,Configuration,Accounting,Performance,Security（FCAPS）functionalmodule.Tested items should inc lude topology d iscovery,service creation,fault monitoring,and performance management.During the test,any created service should be verified w ith emulation service flows.

The OAM mechanism testshould be based on IEEE 802.3ah［9］,IEEE 802.1ag［10］,ITU-TY.1731［11］and G.8114［12］standards.With a p rotocol analyzer,the testis used to check the formats ofContinuity Check（CC）,Loopback（LB）and Link Trace（LT）messages as wellas themessaging flow of Ethernet/T-MPLSOAM mechanism for consistency,seeing if they are comp liant w ith related standards.

3 Practice of CE Test Methodologies

In recentyears,the authorhas participated in the tests of four CE solutions,i.e.PBB/PBT,T-MPLS,Enhanced Ethernetand VPLS using the above testmethodologies.The test results show thatexisting CE solutions canmeet the carrier-c lass networking requirements in term s ofservice bearer capability,reliability,scalability,QoSand OAM mechanism,but they cannot interwork or interoperate w ith each other in the short term because of the g reat differences between them.

3.1 Service Bearer Capability

Test results demonstrate that the abovementioned foursolutions can bear broadband Internetaccess and IP voice services.Butas to p rivate line services,multicastservices and TDM services,their performance and comp lexity in im p lementation d iffer.In bearing p rivate line services,the enhanced Ethernet technology performs poorly if isolated from user information because the entire network devices have to learn the users'MAC add resses.As a result,the network is vulnerab le to attacks.On the contrary,in PBB/PBT,T-MPLS and VPLS technologies,only the UserNetwork Interface（UNI）ofProvider Edge（PE）rather than the core device is used to learn the users'MAC add resses,so the isolation effec tis good.As formulticast services,allfour solutions can deliver the services,but their imp lementation com p lexities d iffer:The enhanced Ethernet technology can easily achieve multicastservices as ring networking mode is adop ted;while the im p lementation ofmulticastservices w ith PBB/PBT,T-MPLS,VPLS/Ethernetover MPLS（EoMPLS）technologies is quite com p licated and diversified and there is stillno uniform standard.For TDM services,PBB/PBT,T-MPLSand VPLS technologies can p rovide TDM interface orCES interface,while enhanced Ethernet technology cannotp rovide full support forsuch services.As to synchronous c lock transfer function in TDM services,mostmanufacturers are trying to develop solutions.

3.2 Reliability

Forb roadband Internetaccess,IP voice services and p rivate line services,allfour solutions can achieve a p rotec tion sw itching of less than 50m s in case of link ornode failure.Formulticast services,some technologies cannot ensure a p rotection sw itching of less than 50ms,which willbe a focus in future research.Moreover,manufac turers adop tdifferent im p lementation mechanism s foraccess link p rotection.In addition to standard Rapid Spanning Tree Protocol（RSTP）,many manufacturers have developed special technologies to imp rove the p rotec tion sw itching efficiency.The test results show thatin case ofa failure,up link traffic w ith RSTP technologymay suffera sw itching over100m s;while the special technologies developed by manufacturers can ensure a sw itching less than 50ms.However,these special technologies are unlikely to interoperate w ith each otherbecause they are p rop rietary to d ifferentmanufacturers.

3.3 Scalability

Currently,the leading CEmanufacturers have developed devices thatsupport 10GE interface,expand ing their interfaces from FE to 10G,thusmeeting the basic requirementofMAN networking.In term s ofsupport for tunnel,the devices,having d ifferentp rocessing capabilities,p rovide differentnumberof services,which range from severalto ten thousands.This canmeet the MAN's m inimum requirements for two-layer networking.Among alltechnologies,PBT,T-MPLS and VPLS are themost scalab le.

3.4 QoS

With respect to QoS flow c lassification and queue dispatching,four solutions can c lassify the flows based on VLAN labeland port,and supportsuch queue d ispatching algorithms as PQ,WFQ and WRR.In addition,mostmanufacturers'devices can im p lement layered QoS function.

3.5 Service Management

Atp resent,manymanufac turers'networkmanagementsystem s can p rovide visualized topologicalviews and enab le service creation but their configuration com p lexities differ.Inmost of these system s,each network element has to be configured separately,thus leading to quite com p licated operations.Only a few of them allow sim p le and quick service creation.In these systems,only the head node and end node ofa service have to be configured during service creation,while the intermediate nodes can be automatically generated and b roadcastby the system.Because the controlp lanes ofPBB/PBTand T-MPLS solutions do notsupportcontrol signals,static configurationmethod is app lied in the networkmanagement system forservice creation.In order to enab le batch configuration ofservices,mostmanufacturers p rovide related App lication Prog ramm ing Interfaces（APIs）and achieve such function w ith sc rip ts.This function cannotbe performed via the networkmanagement interface at themoment,b ringing inconvenience for the operators in daily maintenance.

In respectofOAM support,most manufacturers'devices support802.1ag p rotocol,but the performance management functions defined in Y.1731 p rotocolare only realized in the network managementsystems ofsome manufac turers.

4 Conclusion

As each CE technology has its own app lication scenarios,it is required to make the testscheme based on the technology under testaswellas app lication scenarios.

With increasing attention and investments from operators,manufacturers and research organizations,various CE technologies are continuously developed.Therefore,the testmethodologies should be imp roved accord ing ly so as tomeet the grow ing evaluation requirements for CE technologies.