Metro Ethernet saves a packet

By Nan Chen, Atrica

Enterprise offices are connected internally with 100Mbit/s and Gigabit Ethernet connections and data rides between cities at Terabit speeds on DWDM bundles of 10Gbit/s connections. But the metro area network (MAN) that connects the Gigabit enterprise office to the Terabit network core runs on TDM circuits 64kbit/s to 1.544Mbit/s. The MAN connection also consumes more of the enterprise-networking budget than either the campus network or the intercity core.

Service providers are equally handicapped by existing MANs. ILECs report tens of thousands of held service orders for private-line circuits at T1 and higher data rates. They recognise the complexity of selling, managing and provisioning TDM circuits as a root cause of high operating cost and poor service response. Advanced service offers such as IP-VPN, Internet, ATM and Frame-Relay ride on TDM transport services and are constrained by limited high-cost metro-area bandwidth.

Optical Ethernet can make it profitable to offer the desired bandwidth at an affordable price, since the enterprise then spends more on datacoms. The service provider's capital and operating expenditures are also reduced.

The economics are shown by developing a business case for deploying a suite of Ethernet-based enterprise service offerings based on the geography and demographics of large metro areas, giving operating income, capital and operating cash-flow analyses for each metro alternative: optical Ethernet, Ethernet/DWDM, and next-gen SDH.

Measuring the economic impact of optical Ethernet can yield compelling evidence for adoption in MANs. This includes estimates of capital expenditures, direct operating costs, and other operating expenses. Cash outlays can be estimated by investigating current market pricing and discounting trends, as well as direct and ancillary operating expenses to generate a total cost of ownership measurement. The following cash outlays are estimated:

• CapEx: mainly for optical Ethernet products at the customer's site and in the MAN provider's PoPs.
• Direct Opex: directly related to the Ethernet MAN, including on-going network operations expense, support cost and hardware maintenance expense, environmental expense and fibre-optic facility leases.
• Other OpEx includes marketing and sales, administration and Internet connection fees.

Cash outlays are estimated via a traffic projection developed with an Ethernet service revenue projection and combined with a hypothetical network topology, then used to estimate equipment and facility requirements.

The Ethernet MAN consists of a 10Gbit/s Ethernet physical ring with a full logical mesh core and Gigabit or 10 Gigabit Ethernet subtending rings or spurs. The number and type of customer and POP interfaces, bandwidth required on each ring and within each POP, and number of fibre facilities are all developed from the above establishment-by-establishment modelling process. Network requirements are defined down to each establishment interface and then aggregated to determine the CLE and POP equipment needs. These determine OpEx.

An optical Ethernet edge switch installed at customer sites offers Fast Ethernet, T1 and Gigabit Ethernet interfaces connected on Gigabit Ethernet rings for less than USD20,000 per multi-tenant unit (MTU). This permits service initiation at a price point comparable to legacy T1 service installations but scales to 1Gbit/s at very low incremental cost.

Optical Ethernet core switches installed in the PoPs support both Gigabit and 10 Gigabit optical rings and feature integral optical transport (DWDM). They can scale for rapid service demand growth across a large metro area.

QoS capabilities support ANSI-standard T1 service offers and CIR (Committed Information Rate) and EIR (Excess Information Rate) Ethernet services. This supports rapid service definition and deployment and eliminates truck rolls and overlay networks.

• CapEx, over a four-year period, including the capitalised cost of planning, installation and sparing, is no more than USD18m per year (Fig 1). This supports a revenue base of over USD100m in the first year and USD700m in the fourth year.

Core equipment is highly cost efficient due to integration of advanced Ethernet switching functions and optical transport.

• Direct OpEx is mostly metro access fibre leases (Fig 2), if fibre-optic facilities are treated as lease expenses. Metro core fibre lease expense is negligible because the core switch provides multiple 10 Gigabit Ethernet wavelengths on a single fibre using integrated Ethernet switching/DWDM.

On-going OpEx (capacity planning, network administration, customer care) is the second-highest direct OpEx, but — through system integration and software-controlled provisioning/management — is <4% of year-4 revenue, compared to 30% using legacy technology.

Environmental and equipment support costs — determined from the number of shelves of each equipment type required and then computing resource costs from data-sheets — can be very low due to the density of equipment.

• Other OpEx items tends to scale with revenue. Efficient use of CapEx and direct OpEx leaves the headroom to aggressively grow business through ambitious sales and marketing.

Revenue is projected by estimating take-up rate and usage of the Ethernet service suite by enterprise establishments in a large metro area. The addressable market and distribution of establishments is determined by combining demographic data on the type, mix and size of establishments with a hypothetical four-year phased build-out of metro optical rings. Four types of customer sites are modelled: MTUs (including large, medium and small tenants); medium enterprise sites; large enterprise sites; and service provider sites. Service revenue is then estimated recursively as follows.

1. The number of establishments that subscribe for any element of the service suite in their first year of eligibility is calculated by applying a take-up rate for each establishment type.
2. A usage profile for each subscriber by establishment type during year 1 is used to estimate which service and options will be purchased.
3. Estimate bandwidth usage by establishment type, service offer and option, for each year.
4. The service pricing curves in the preceding section are applied to the usage per subscriber amounts of Step 3 to provide revenue by establishment type, by service and service option, and by year.
5. This process is repeated for years 2, 3, and 4. This recognises usage differences according to when establishments first become subscribers.

The four-year revenue forecasts by establishment type are based on modelling demand for a full suite of highly flexible and scalable metro Ethernet services that meets enterprise private network and Internet communications needs. Its five service offerings include:

• Ethernet Internet Access Service
• Ethernet Transparent LAN Service
• TDM Service
• Metro Ethernet Private Line Service
• Custom Dedicated Ethernet Private Network.

The suite is an extension of existing service offerings and can be used by an incumbent to develop new sources of revenue and profits from existing customers while avoiding cannibalisation of existing services.

Ethernet services also provide a way to build customer loyalty and reduce churn by meeting forward-looking customer requirements. Ethernet services can also incrementally increase customer usage through promotions and differentiated pricing plans because bandwidth and QoS modifications can be made in arbitrarily small increments, unlike TDM and frame relay services' hard breakpoints at 64kbit/s, 1.5Mbit/s, 45Mbit/s, 155Mbit/s etc. This also cuts customer retention and churn by eliminating big decisions to upgrade to the next bandwidth.

Comparison for total cost of ownership (TCO) with two other architectures — a traditional Ethernet switch coupled with WDM transport and next-generation SONET (using network designs that meet the above service demands and topology) — show a substantial CapEx and OpEx advantage.

In the Ethernet switch/WDM transmission system solution, several sizes of Ethernet switches are used, including large and small enterprise switches for use as CLE and carrier-class models for switching within the POPs. The WDM transport equipment is protocol neutral. Separate Ethernet switches and WDM transport systems yield a significant TCO disadvantage. Also, increased design complexity gives poorer reliability and lack of network flexibility and responsiveness.

Optical Ethernet out-performs the alternatives by more than 4:1 on ROI (Fig 4), moreso as it scales to Terabit usage on the core network. Using the same revenue model and adding estimated CapEx and OpEx for the competing architectures shows that optical Ethernet enables build-out a highly profitable metro Ethernet service on a large scale. Also, the TCO comparison shows that it is much less costly in both CapEx and OpEx. Specifically:

1. An optical Ethernet MAN supports robust revenue growth while rapidly achieving high returns. Year-1 revenue is USD150m and year-4 revenue USD650m, with a sustainable ROI of 400% by Q2 of year-1.
2. Optical Ethernet pays back its cost by Q3 of year-1 versus 2 years for Ethernet/DWDM and 3.5 years for next-gen SDH.
3. Optical Ethernet TCO is 43% less than Ethernet/DWDM; 51% less than next-gen SDH — less costly for each CapEx and OpEx element due to integration of Ethernet switching and optical transport in both CLE and POP products and scaling to Terabit capacity.

Nan Chen
Product Marketing Director
Atrica
[email protected]

Chen is the founding president and a board member of the Metro Ethernet Forum and was a founding director of the 10 Gigabit Ethernet Alliance.