Why should you select EoMPLS as the technology of choice for a green field CE deployment?

Introduction

Assume a Communication Service Provider (CSP) who does not have an aggregation network between the access and core. The CSP wants to use Carrier Ethernet (CE) technology to build the aggregation network. The CSP, most of the CSPs, intends to deliver/transport/aggregate the following services/traffic across the Carrier Ethernet Network (CEN) in an End-to-End (E2E) Internet protocol (IP) oriented architecture supporting the current and future demands;

o   Enterprise and corporate data services

§  Layer 3 Virtual Private Networks (VPN)

§  L2 VPN

§  L2 Point to Point (PP)

o   Consumer services

§  IP Television (IPTV) (multicast TV and unicast Video on Demand (VoD))

§  Broadband Internet

§  IP Voice

o   Wholesale services

§  Mobile backhauling

§  L2 VPN

§  L2 PP

IP/Multi-Protocol Label Switching (MPLS) has been a field proven technology and assume, like most of the CSPs, this CSP also has for the past several years implemented an IP/MPLS based core network and has built a team of experienced staff.

Options

To deploy a new CEN technology, the CSP has the following technology options available.
 

Feature	Provider Bridging (PB) / QinQ / Institute of Electrical and Electronic Engineers (IEEE) 802.1ad	Provider Backbone Bridging (PBB) / MAC in MAC / IEEE 802.1ah	PBB-Traffic Engineering (PBB-TE) / IEEE 802.1Qay	MPLS-Transport Profile (MPLS-TP)	IP/MPLS	Ethernet over Synchronous Digital Hierarchy (EoSDH)
Control Plane			Centralized Server	Centralized Server	MPLS	Centralized Server
Provisioning	Command Line Interface (CLI),  Element Management System (EMS)	CLI, EMS	EMS/Network Management System (NMS) only	EMS/NMS only	CLI, EMS	EMS/NMS only
Multipoint support	Yes	Yes	Yes	Yes (with VPLS)	Yes	No
Media Access Control (MAC) learning	CSP needs to learn all the MACs	reduced MAC learning	Not automatic	Yes (with VPLS)	Yes (Virtual Private Local Area Network (VPLS)	N/A
Protection	Spanning Tree Protocol (xSTP)	xSTP	Ethernet Ring Protection Scheme (ERPS)	Not mature Note 1	50 millisecond (ms), ring/mesh	50ms, ring
Operation			SDH like	SDH like		SDH like
Addressing scheme	Ethernet	Ethernet	Ethernet	different	Ethernet	N/A
TE	Poor	Poor	Good	Good	Good	Poor

Note 1: RFC 6372 (MPLS-TP Survivability Framework) released on September 2011

Feature	PB / QinQ / IEEE 802.1ad	PBB / MAC in MAC / IEEE 802.1ah	PBB-TE / IEEE 802.1Qay	MPLS-TP	EoMPLS	EoSDH
Maturity	High	Low	Low	Low	High	High
Interoperability	High	Low	Low	Low	High	High
Scalability	Low	High	High	High	High	Low
Separation of customer networks	High	High	High	High	High	High
Vendor stickiness	Low	High	High	High	Low	Low

Standard Defining Organizations have already published several standards for the IP/MPLS based CENs considering the current and future IP oriented service requirements. Following are some of the recent standards;

o   Broadband forum’s TR-221 (Technical Specifications for MPLS in Mobile Backhaul Networks)

o   Broadband forum’s,  WT-224 (MPLS in Carrier Ethernet Networks)

IP/MPLS in the CEN is required to support IP VPN and IP multicast features to support 4^th Generation (4G) mobile services such as Long Term Evolution (LTE) (all IP architecture).

To deliver the above mentioned multiple services on a single CEN with required service features, IP/MPLS is the most suited and matured technology. It is also needed to inter-op with the CSPs existing IP/MPLS core, especially for E2E seamless services.

It’s also noted that most of the access network uplinks are Ethernet or becoming Ethernet, while SDH/Plesiochronous Digital Hierarchy (PDH) networks are becoming outdated and obsolete. Hence investing on SDH is pointless.

Recommendations

Among many others, following are the most important recommendation for the CSP.

• The CEN shall be Transmission Agnostic.
• Aggregation of topology - to reduce the numbers of physical interfaces required at higher levels of the transport / switching hierarchy.
• Consolidation of network and transport protocols - to reduce the complexity of logical interfaces required at higher levels of the transport / switching hierarchy.
• To have different traffic types physically and logically aggregated, so that they can be transported by the Core Network.

• Development of CEN shall be closely mapped to service development strategy.

• The CEN shall be EoMPLS based.
• When selecting an CEN site, following shall be considered;
• number of access nodes in proposed topology
• total bandwidth demand and future bandwidth forecast that has to be aggregated and carried
• cost of alternate aggregation options and possible service impacts in a failure or other adverse condition
• All the network elements of different switching capacity and network shall have high availability features.
• The service delivery architecture within CEN shall be layer 2 based except for multicasting which shall be IP multicasting. However layer 2 multicasting features shall be available for customer multicasting services.
• The services and equipment shall be certified with Metro Ethernet Forum (MEF).
• Virtual Routing and Forwarding instance (VRF) shall not be brought to CEN level, unless it’s required for 4G Radio Access Network (RAN) backhauling (LTE) in the future as specified in Broadband forum TR-221.

CEN Requirements

Any CEN architecture needs to support key requirements of availability, stability, Quality of Service (QoS), performance, multicast support, Time Division Multiplexing (TDM) support, management and security. In the context of an EoMPLS based CEN,  that translates to what is explained below.

Availability (Resilience)

Since the CEN interfaces with the access layer, the resiliency is a key factor to avoid service outage due to the node failure or link failures. This is achieved by adopting multi-homing topology for the interconnection between CEN and the service edge network as well as the CEN and access layer where feasible. However, in order to avoid the complexity of the CEN, not more than two connections towards service edge network is recommended. Network side connections used for multi-homing requirements shall be terminated on physically separated line modules.

Equipment/node level high availability shall also be employed to ensure service and network availability due to failure of critical hardware and software modules of the CE node. The following redundancy mechanisms shall be available in all the network elements of CEN.

Hardware Component of CE Node	High Availability  mechanism
Route processor	1:1
Switching fabric	1:1
Power supply	1+1 Note 2
Power feed	1+1 Note 2
Cooling system	1+1 Note 2
Any other control plane module	1:1
Any other switching plane component	1:1

Note 2:   single component shall be able to take the full load of the CE node

The network shall use International Telecommunication Union-Standardization (ITU-T) G.8032 version 1 & 2 (ERPS) wherever possible to achieve sub 50ms protection and recovery for Ethernet in ring topology in case of a node or network failure.

Following software level high availability features shall be implemented;

• Non Stop Routing (NSR) for
• Label Distribution Protocol (LDP)
• Resource reservation Protocol (RSVP) TE
• Border Gateway Protocol (BGP)
• Open Shortest Path First (OSPF)
• Protocol Independent Multicast-Sparse Mode  (PIM-SM) and PIM-Source Specific Multicast (SSM)
• In Service Software Upgrade (ISSU)

Bidirectional Forwarding Detection (BFD) requirement shall be analyzed for following in the future stage

• LDP
• RSVP
• BGP
• OSPF
• PIM-SM and PIM-SSM

Non Stop Forwarding (NSF) requirement shall be analyzed in future stage.

Stability

Stability of the CEN and its Network Elements (NE) are very important. This should ensure consistent performance of the NE. Mean Time Between Failures (MTBF) and Mean Time To Repair (MTTR) values shall meet 99.999% node availability requirements.

QoS

CE QoS model is essential to provide dynamic quality of service feature without overbooking the bandwidth for applications. It should be able provide better service to selected traffic, depending on the individual requirements of different types of service and also to meet requirements in customer Service Level Agreements (SLA).

The CEN shall be able to handle Layer 2 (802.1Q), Layer 3 (Differential Services Code Point (DSCP)) and MPLS (Experiment (EXP)) QoS/Class of Service (CoS).  The CEN/NEss need to support classifying, marking, remarking, scheduling, shaping and policing for all the above QoS/CoS models at all egress and ingress ports whether access side or network side. Within the CEN, the nodes shall be able to support at least 4 hardware queues for traffic per port. The control and management traffic within the node shall be handled separately from the user traffic. Hierarchical QoS shall be analyzed in future stages.

Performance

Scalability of the CEN determine by providing sufficient bandwidth to be able to guarantee a committed level of performance for the full service portfolio of end users. CEN design to achieve the certain QoS requirements/Key Performance Indicators (KPIs) defined with the set of services/products. The CEN must be able to handle unpredictable surges in traffic, and appropriate load. The network utilization has to be maintained within 70% to facilitate the introduction of services and for the unpredictable surges in traffic.

Multicasting

To support IPTV and other multicast applications, the CEN shall support IP multicast protocols.  Layer 3 based (PIM) multicast technology is preferred over Layer 2 technology for scalability and flexibility reasons. Layer 2 multicasting features shall be available for customer multicasting services.

Supporting TDM services

TDM services shall be supported in the form of Circuit Emulation Services (CES) using Synchronous Ethernet (EtherSync/SyncE) or IEEE 1588v2 for frequency and time of day synchronization. At least E1 and STM-1 CES shall be supported. Enabling CES shall be done if the no. of TDM services few compared to the Ethernet services. If not, separate TDM equipment shall be used.

Management

All the CE elements should be able to address the management domain requirements. The standard functional entities such as: Fault management, Configuration management (Fulfillment support), Security management, Performance management and Inventory management on all Network Nodes will be required.

For service management, the network and the nodes shall support following Ethernet Operation Administration and Maintenance (OAM) standards;

• IEEE 802.1ag (Connectivity Fault Management (CFM))
• IEEE 802.3ah (Ethernet in the First Mile (EFM))

Security

The CEN addresses the security which provides confidentiality, integrity and availability of specific services. The following areas have been identified and will be equipped with necessary security mechanisms,

·         Node security

·         Access security

·         Interconnection security – User to Network Interface (UNI) and Internal-Network to Network Interface (I-NNI)

·         Protocol security – UNI and I-NNI

Standardization

MEF, the Broadband Forum, Internet Engineering Task Force (IETF), IEEE and ITU-T are the main Standard Defining Organization (SDO) with regard to the CEN.

The EoMPLS is a field proven and matured technology in implementing CENs. Though the standards are available, CSP needs to standardize this architecture and protocols. All the future network developments and deployments in the future need to align to these.

It is recommended that CSP get involved with these SDOs, especially the MEF and the Broadband forum. It is also recommended that CSP get the MEF certification for its services (E-Line, E-LAN, E-Tree and E-Access) and use MEF compliant equipment in the CEN (MEF 9- Ethernet Services at the UNI, MEF 14- Traffic Management Phase 1). Click here to see the 5 CE attributes defined by MEF.

When and why you should move to Carrier Ethernet?

First things first, I would suggest the Communication Service Provider (CSP) should critically analyze the book value of the existing legacy system, spare capacity availability, vendor support availability, Next Generation Network (NGN)/Ethernet transport feature availability, upgrade costs and operational costs like space and power. As long as the existing system addresses the requirement cost effectively in short term and mid term, I would suggest that the CSP continue with the legacy system with a clear replacement/migration plan for long term.

I'll now list 2 important highlights below to show you why I would replace an existing legacy (ex:- SDH) network with a Carrier Ethernet (CE) network.

(1) *Native* support of cost effective and ubiquitous Ethernet interface/service/transport

As you know the current and future traffic on CSPs are more towards IP/Ethernet and this requires that CSP
1. provide "Ethernet interfaces" to customers,
2. transport/provide "Ethernet Services" like MEF E-LINE, E-LAN, E-TREE and E-Access,
3. build "Ethernet based transport networks" like Carrier/Metro Ethernet MANs/WANs.

If you are to transport Ethernet over SDH, you require EoSDH capabilities on your SDH platforms. On the other hand, for you to transport TDM over Ethernet, you require Circuit Emulation Service over Packet Switched Networks (CESoPSN)/Structure Agnostic Transport over Packet (SAToP). Since the trend is now to move from legacy (TDM) to NGN (IP/Ethernet), you will see more IP/Ethernet than TDM in the future. So it's logical to have CE with TDMoEthernet, rather than having SDH with EoSDH. Finally all will converge to CE.

(2) Vendor support

As these trends happen in the CSP ecosystem (CSPs, OEMs, SDOs etc.), the vendors are also discontinuing legacy systems and move to NGN based technologies like CE. This will make the cost of legacy systems high and cost of NGN systems low due to the economies of scale. This will also made the support (AMC- Annual Maintenance Contract) charges for legacy systems be also high.

National Broadband Network reality

I think the operators/governments, both developed and developing, around the globe need to go back to the fundamentals and 1st principals in deploying broadband/Next Generation Access (NGA) networks. For that matter, most governments need to go back to fundamentals/1st principles in all that they do.

They need to think on the "purpose" of doing what they are doing/planning to do. If you continuously search for the purpose of the purpose and so on, then only you'll find what is actually required to be done. For example, take broadband. Why broadband? to have a high bandwidth delivered to homes. Why high bandwidth? to deliver more content/applications. Why more content/applications? for information (ex: Internet) and entertainment (ex: IPTV) delivery. Why information? to make the life convenient.

Ok. now, to make the life convenient, what else can we do other than providing broadband? I think the governments can do lot more things together with broadband to make peoples/citizens lives convenient. ex: good governance, public reforms etc. Is Fiber to the Home (FTTH) the only solution? The governments/operators need to broaden their solution space and take in to account all the possibilities and find the optimum solution, without being fascinated and misdirected by the hype. There's no need for you to follow, just because someone has done it.

Nothing is complex; you need right thinking, attitude and desire. If governments got that together with common sense and looking for purpose of all they do, they can.

Broadband: wired or wireless?

I always had this question in my mind; wired or wireless to reach the rural.

All the fixed line Communication Service Providers (CSPs) in both developing and developed countries have started their networks form the urban areas (ex: Main Switching Unit (MSU) in Public Switched Telephone Network (PSTN)) and move to sub-urban (ex: Remote Switching Unit (RSU)) and then to rural (ex: Line Extension Unit (LEU)). For both copper and fiber, this has been the practice for the last 150 years. Even wireless CSPs have followed the same path first covering the urban, then sub-urban and then rural. Two of the most obvious reasons for this, is the demand and the affordability.

When it comes to broadband (BB), now we have 2 options wired or wireless. While there are methods to re-arrange an existing copper network by shortening the copper distance and introducing FTTN/C/B to deliver BB using ADSL2+/VDSL, the re-arrangement of wireless networks are not that feasible. To cover the green areas, the operators need to deploy new networks; copper, fiber or wireless.

As you move from the urban to rural, the population/house-hold density decreases, making the requirement of coverage large. The 2 options remain as fiber and wireless (ex: LTE). Deploying new copper is not a choice any more.

But the question is, once you reach the rural, both demand and affordability decreases, especially in developing countries. On the other side, the requirements are becoming more scattered in nature making wired networks prohibitive. Therefore the ideal choice is wireless, because of the less demand, less affordability and scattered nature.

So, the viability of FTTH is only applicable in urban/sub-urban areas where the required bandwidth is not deliverable even with the existing copper network re-arrangement as described above or the copper is outdated/degraded/old. For other areas (rural) wireless is the only viable options, especially for a developing country. Please keep in mind that the green areas will most of the time are rural.

For a case study, I'll take my own country as an example. The population of Sri Lanka, a developing country, as at March 2012 is 20.3 million (approx.) [1]. According to [2] 78% of the population lives in rural areas. If we assume a house hold has 4 people, the number of house-holds is 5 million (approx.). 78% of this is 4 Million (approx.).

But, according to countries regulator [3], as at March 2012, the number of fixed lines in Sri Lanka (including CDMA) is 3.6 million. However, we need to keep in mind that big corporates use hundreds of lines per location. Therefore, we cannot use 3.6 million for the calculation.

The point however is, 78% of the households/population are rural. To reach this we have to go with wireless. This is quite evident from the figures in [3], where the mobile penetration is 91% while the fixed line penetration is 18% (approx.). This 18% is actually the saturated figure of urban and sub-urban. But when the rural areas develop (when the demand and affordability increases) the 18% will increase as explained in my previous post. In [3], the Internet penetration is also high with mobile (if calculated, 4% approx.), whereas for fixed it's only 2%, if calculated. Obviously the BB is also will follow the same percentages.

On the other hand, the 78% includes large number of Small and Medium Enterprises (SMEs), which are the main contributors to the country’s economy. For them to grow and the country to grow, the SMEs need Information Communication Technology (ICT) facilities. So the operators and the country have a great challenge to provide them the required services. Operators need to invest cautiously making sure they get positive Return-On-Investment (ROI) at the same time increase the footprint of their service coverage.

[1] http://www.statistics.gov.lk/
[2] 202.11.2.113/SEBM/ronso/no3_4/aruna.pdf
[3] http://www.trc.gov.lk/images/docs/statis_March_2012.doc

E-LAN vs. E-TREE

Metro Ethernet Forum (MEF) has defined 4 types of service, namely;

1. E-LINE (Ethernet Line) - Point to Point
2. E-LAN (Ethernet Local Area Network (LAN)) - Multipoint to Multipoint
3. E-TREE (Ethernet Tree) - Point to Multipoint
4. E-Access (Ethernet Access)

The lack of E-TREE implementations/support on Communication Service Providers (CSP) networks today make the customers buy Virtual Private LAN Service (VPLS) (Multi Protocol Label Switching (MPLS) bases E-LAN implementation) services for point-to-multipoint applications such as 3G back-hauling to Radio Network Controller (RNC), Head Quarters (HQ) to branch communication on corporates. 

If the resources required to E-LAN and E-TREE are compared, when the E-TREE requires N number of Pseudo Wires (PW)s for N locations, E-LAN requires N*(N-1)/2=> N^2. If the Media Access Control (MAC) learning is considered, in E-TREE leafs learn only the root MACs and root learns all the root and leaf MACs (2 if there's only one root exists). In E-LAN, all the points in the VPLS (Virtual Switching Instance (VSI)) will learn all the MACs (N). 

Note that MAC table size is a limited resource for a Network Element (NE). Number of PWs, number of Label Switch Path (LSP)s, number of VSI instances and number of Virtual LAN (VLAN)s are also limited resources.

Next Generation Optical Transport Networks

When we say Optical Transport Networks (OTN), it could mean two things;

1. OTN wrapper capability 
2. OTN switching capability

To implement an OTN, there are many technological options available. From least cost (in general) to highest, the options are as follows;
Note: All options need to support OTN wrapper

1. Fixed Optical Add Drop Multiplexers (FOADM)
2. Re-configurable OADM (ROADM)
3. Tunable ROADM (TROADM) (Wave-length Selective Switching (WSS)): supports Color-Less and Direction-less)
4. FOADM with Automatically Switched Optical Networks (ASON)/ Generalized Multi-Protocol Label Switching (GMPLS) control plane
5. ROADM with ASON/GMPLS control plane
6. TROADM with ASON/GMPLS control plane
7. FOADM with ASON/GMPLS control plane and OTN switching
8. ROADM with ASON/GMPLS control plane and OTN switching
9. TROADM with ASON/GMPLS control plane and OTN switching

Other than Color-Less and Direction-less, Contention-less is also a good feature to have on WSS systems. 

Cost of adding OTN switching capability vs. loosing sub-lambda grooming at intermediate sites need to be properly analyzed based on your current and future traffic matrix. 

To tell you the truth, OTN switching is more a hype than a reality. This is quite evident from the low number of OTN switching deployments currently in the world.

The prime advantage of OTN switching is the sub-lambda grooming at intermediate sites. The industry trend(both suppliers and operators) is to start without OTN switching and go for OTN switching in the future if all the lambdas run out/close to run out (aka Wave-length blocking). This requires that you select a vendor who's capable of OTN switching but you need not purchase OTN switching components (cards) on day one.

You do not need OTN switching to achieve mesh protection. What is then required is ASON/GMPLS.

A good approach,adopted by many operators when publishing Request For Proposal (RFP)s for Optical Transport Networks is keeping the RFP open for all the options given above. It's required to give the fiber characteristics, locations and the traffic matrix (current and future). Based ion these inputs the vendors can come out with the least TCO option. The evaluation should also be based on lowest TCO (this covers both initial Capital Expenditure (CAPEX), future expansion CAPEX and the running Operational Expenditure (OPEX) such as site rentals, power etc.).  

When you want to do sub-lambda grooming at intermediate sites, you'll have to have OTN switching (CAPEX!).

When you have OTN switching, the earlier Point-to-Point lambda passed through several intermediate nodes at the optical domain (OOO) now need to go to electrical domain to do grooming (OEO) making it multi-segment. This requires several OTN ports (CAPEX!). However, you use only one lambda. Some call the latter as Layer 1-ASON and former as Layer 0-ASON.

If you do not do sub-lambda grooming at the intermediate site, you will have to have a separate lambda (CAPEX!) at the intermediate site, though the traffic goes to the same destination.

The above 3 CAPEX components need to be properly analyzed for the current and future traffic matrix. Then only the most optimized design and then the most optimized cost can be calculated.

For the comparison, following options are recommended to be used as ASON/GMPLS is better to have. This will also make sure that the comparison is more balanced.

1. FOADM with ASON/GMPLS control plane
2. ROADM with ASON/GMPLS control plane
3. TROADM with ASON/GMPLS control plane
4. FOADM with ASON/GMPLS control plane and OTN switching
5. ROADM with ASON/GMPLS control plane and OTN switching
6. TROADM with ASON/GMPLS control plane and OTN switching 

"Digital Optical Networking (DON)" technology claimed to be available with some vendors can address the sub-lambda grooming at optical level without using OTN switching (i.e using OOO as opposed to OEO) giving cost advantages.The component used is known as "Digital ROADM".

 

Future of wire-line access networks

Introduction

Communication Service Provider (CSP) networks can be divided into many categories. If divided based on the media used for the transport, wired and wireless are the widely known. Both types can be again sub divided based on the functional hierarchy; core, aggregation and access. If wired networks are concerned, optical fiber is used in almost all the core and aggregation networks because of its ability to carry/transport large amounts of data.

Wire line CSP access networks can be built using cable, copper or optical fiber. The cable based access networks are common in North America. Countries like Sri Lanka are more used to copper and fiber based access networks.

Access networks can again be classified based on the customer segment. While it is simple to have a single access network for both residential and business customers, some CSPs build different access networks for these 2 segments. The requirements and expectations of the 2 segments are different.

Design Criteria

A best practice in any access network design is to start with some forecasted marketing data. The networks that we build today should be future proof as far as possible in terms of service demand and technology used. For an example, for a residential triple play access network, it is better to have some figure on the number of customers per area, type of services required by those customers, the bandwidths consumed for each service and the bandwidth per house hold. Based on this information, we can find out the immediate bandwidth requirement per customer and the future demand. While there are many differences between fiber and copper based access networks, the fundamental difference is the ability of fiber to carry large amounts of data. Optical fiber networks in that sense are very much future proof.

It is always a good idea to design the networks for your real requirement rather than going behind various technologies. A technology used in United States (US) or Europe may not be able to directly deployed in countries like Sri Lanka.

The main decision factors when deciding on an access network technology/media are cost per user, bandwidth demand and demographic criteria. The current trend is to deliver more bandwidth per user in a secure way at lower cost.

Copper or fiber

If copper is already laid and available to the households, then it will be a good idea to use them if the required bandwidths can be delivered using them. The fundamental problem of copper based access is the decaying of speed with the distance. Most of today’s copper based access networks start from the Central Office (CO) of the CSP and goes to the customer. The access network is fully passive, consisting of, but not limited to, Main Distribution Frame (MDF), primary cable, Cabinet, secondary cable, Distribution Point (DP) and overhead cable. The popular Digital Subscriber Line (DSL) technologies are struggling to deliver high bandwidths when the distance between the CO and the customer increases.

One solution is to use optical fiber, instead of copper. But the main problem of copper, as identified earlier, is the decaying of speed with distance. While it is accepted that the ideal solution is to have fiber, we can also try to shorten the copper length and deliver high bandwidths. This will save lot of cost as fiber is not freshly drawn.  This introduces a new term called FTTX, Fiber To The “place you want”. X could be building, home, curb/cabinet, node or even desk. So now the issue is between FTTH and FTTC/B.

Optical access networks

Optical access networks could be active or passive. An example of an Active Optical Network (AON) is Metro Ethernet or Carrier Ethernet. Here, active Ethernet switches are deployed in the network to deliver services, mainly to business customers. The networks mostly take the form of rings assuring high availability required by business critical applications.

The other category is Passive Optical Networks (PON).  PON works by delivering an end to end fiber access to the building or home. Though PON can be used for FTTC/N applications, most PON applications are based on FTTH/B architecture. Unlike AON, PONs have limited network protection in the last mile, because of its passive nature. PON works by dividing an optical signal into multiple fibers using a passive optical splitter. After the splitter, the network is liner and does not provide any direct protection. This type of a solution is mainly suitable for a residential rather than business.

The Table 1 below gives a snapshot of various PON technologies.

TDM PON			WDM PON	Hybrid TDM / WDM PON
ATM PON (APON) =Broadband PON(BPON) =A/B PON	Ethernet PON (EPON) = Gigabit Ethernet PON (GEPON )	Gigabit PON (GPON)
ITU-T G.983 standard	IEEE 802.3ah standard	ITU-T G.984 standard		Still at research stage
Developed from Telco side	Developed from Internet side	Evolved from A/B PON	Evolution from TDM PON
Layer 2 encapsulations are Ethernet and Asynchronous Transfer Mode (ATM)	Layer 2 encapsulation is Ethernet	L2 encapsulations are GEM(GPON Encapsulation Method) for Ethernet and ATM
Maximum up stream is 155Mbps, Maximum downstream is 622Mbps	Maximum up stream and downstream is 1.25Gbps	Maximum up stream and downstream is 2.5Gbps	Maximum up stream and downstream is 10 Gbps
Deployments: US (ex: Verizon FiOS)	Deployments: Japan (ex: NTT, KDDI), Korea (ex: KT), China, India	Deployments: US (ex: AT&T), Europe	Deployments: Korea (ex: KT), China

Table 1. PON technology comparison.

TDM PON is the current choice because of its low cost. Out of the available TDM PON technologies, GPON has much better multi-service capabilities and carrier grade management capabilities and therefore the winning technology. It is also future proof, because of its high bandwidth support.