While the focus so far is on delivering 6 Mb/s connections to homes but providing a good user experience will require carrier Ethernet all the way into homes but Malaysia is not yet fibre rich right now, according to Andrew Coward, Juniper Networks vice-president of Service Provider Marketing & Partnerships.

Ethernet is a protocol which defines a packet structure which can be delivered over Synchronous Digital Hierarchy (SDH), Asynchronous Transfer Mode (ATM) or other such legacy networks but it's costly per bit, hence the need to move to a full IP-based network.

“After all, 80% of a carrier's spending is operational expenditure, so if this can be reduced to say 60%, there would be more to spend on development,” said Coward on the sidelines of the recent Carrier Ethernet World Asia Pacific Congress in Kuala Lumpur.

Carrier Ethernet can deliver all legacy services such as VPN, fixed broadband access, mobile voice and wireline voice, as well as new data services such as streaming video, digital TV, managed telepresence, online ads, mobile advertising, location-based services, mobile data services, wireless broadband services, CDN and cloud computing – to name a few – which to be delivered cost effectively must ride on a shared packet based IP network.

These new services are expected to experience 20% CAGR (compound annual growth rate) according to a December 2008 Yankee Group report, while the legacy services mentioned above have very low single digit CAGR at best, so the challenge is to transition the network and the business model to take advantage of  service opportunities and offset declining revenue and margins from more traditional services.

However, Ethernet was designed as a best-effort service within a local-area network (LAN) environment but such shared bandwidth generally results in a poor experiences, especially when viewing streaming videos, so the challenge is to take the cost advantage of Ethernet combined with proper controls to enable high quality service and a good user experience.

"For example, a high-definition conference call won't be possible unless it's carried over private circuits or service providers enable assured bandwidth for its short duration, for which users pay a premium,“ said Coward.

It also requires new service revenue models where telcos must be prepared to share revenue with, as content providers are increasingly forming direct relationships with end users, while service providers simply bill users at the end of each month.

What's also happening is that content providers are forming relationships with service providers to provide a good end-user experience, for example, Comcast in the U.S. which offers 10 to 12 channels on demand.

Two solutions

There are two models which can be adopted. One is a revenue sharing model where if a premium customer plays a game, the service provider gets 20% of the revenue.

For example, the Microsoft Xbox backend gaming server has transactional capabilities with service providers, while Juniper and Polycom have developed a transaction engine where the network sets up the required bandwidth for a telepresence session – i.e. high-definition video conferencing. 

No game service so far brings in as much revenue as voice in many networks and voice switches have a lifespan of 10 year, while a DSLAM (digital subscriber line access multiplexor) used with ADSL has a lifespan of three years and telcos want to amortise their voice equipment, which makes it harder to justify investment in equipment based on newer technologies.

However, once the broadband market reaches saturation, then there is no viable return-on-investment to be had.

For example, an ADSL connection costs around US$15 per month in Japan, while a 100Mb/s connection costs US$25 per month and a 3 Mb/s connection costs US$35 per month in California.

In a voice dominated network, revenue tends to track increases in traffic volume and remains above network cost, thus maintaining profit but in a data dominated network, as traffic increases, revenue dips below network cost, resulting in an overall loss.

Also, if a service provider increases a subscriber's bandwidth from say 3 Mb/s to 10 Mb/s it cannot increase its charge for that higher bandwidth by the same factor, so its profit margin goes down.

So service providers are pressured into providing managed services by working with content providers to provide them with prioritised bandwidth. For example, Polycom is conducting trials in three places worldwide – one of them in Vietnam where the transaction engine in the carrier network signals to the routers to reserve bandwidth for the video call – much like a virtual-private network (VPN) for the duration of the call. Likewise for some IP-TV packets which are prioritised ahead of normal IP traffic.

However, this is rather controversial in terms of net neutrality, since it affects other content providers which don't have such relationships.

Fortunately, the second model, which does not impact upon net neutrality, is for the content provider to pay service providers a flat fee for an online game, IP-TV or other bandwidth demanding application  with more access switches (nodes) in the Carrier Ethernet Transport network, especially in central offices nearest to the end-users.

These nodes as well as aggregation switches in the Carrier Ethernet Transport (CET) network are connected to dense-wave division multiplexing (DWDM) metro optical rings in the optical transport layer below it, while core switches in the CET network connect to the DWDM optical core rings below it, as well as the IP/MPLS (Internet Protocol/Multi-Protocol Label Switching) core which performs the routing functions above it.

The above solution is the result of a joint venture between Juniper and Nokia Siemens Networks concluded in October 2009, which brings together the best of Carrier and Metro Ethernet products – switches, routers and management tools - to deliver converged Transport and IP Carrier Ethernet solutions.

The basis of this solution is a flexible service architecture combing Ethernet with seamless MPLS, to adaptively place the service edges (i.e. - nodes) where needed and to balance between scalability, bandwidth optimisation and bandwidth savings.

A gradual process

However, getting carriers to migrate to an all IP-based network isn't all that easy, especially since they have invested heavily into extensive legacy networks and technologies.

“Carriers tend – as all investors do -  to follow a down-to-earth vision. They plan network evolution on the basis of detailed analysis supported by trials, tests and evaluations. But their near-term investments must take long-term vision into account. This means the evolution to a consolidated network architecture must still protect the investment in existing legacy but must simultaneously enable new packet services that drive revenue and margin,” said Coward.

This leads to the sticky question - “When is the right time to migrate my TDM (Time Division Multiplexing) services into a pure packet network without disturbing my existing revenue flows and without affecting service quality?”

“From experience we can say there is no big-bang approach and those that have talked publicly about an overnight switch-over have found (albeit privately) that a step-by-step approach is the only logical and economically viable approach,” said Coward.

The timing of the migration is based on the assumption that the technology available has the capability to emulate or transport the existing services without loss of quality or user experience.  Advances in the pairing of Ethernet (from a cost and bit-rate perspective) and MPLS (from a reliability and service quality perspective) over the last three years provide today, the necessary attributes to make this transition not just possible but highly desirable and are forming the foundation of service provider’s growth plans. Critically, this technology pairing means that customers are served not just raw, commodity bandwidth, but real service value.

Unfortunately, recent years have seen not consolidation in network infrastructure but parallelism – the build-out of parallel networks to enable new services without turning off or consolidating existing services.

Even as these new services generate top line revenue, the operational cost inefficiencies sink margin. Without associated infrastructure optimisation and network cost reduction, the challenge of this approach is to stay above the cost curve especially as growth of new services actually increases near-term infrastructure costs, e.g. by requiring separate service silos.

So it therefore gets harder before it gets easier - in transition, operating both packet and circuit networks together can create more complexities and higher costs.  But being able to maintain the present mode of operation of transport networks through the transition, eases this migration.

An additional challenge through this consolidation is bringing together the transport view of the world with the IP view.   These have almost been separate industries and in most service providers are represented by different groups.

It turns out that these two groups are no longer poles apart as the transport world view moves to a more dynamic approach to service delivery and the IP world takes on many of the attributes and reliability characteristics of the transport world.   One of the “binding agents” for this has been the development of MPLS-TP which marries much of the transport requirements with a lighter version of the service-rich MPLS feature-set.

The first fruits of this unified Carrier Ethernet solution for converged networks by the Juniper/Nokia Siemens Networks joint venture will be introduced in the first quarter of 2010,

Built on Carrier Ethernet Transport. The benefits of this approach include the cost-effectiveness a simple access technology provides, a connection-oriented solution that gives complete control over the network with no spoofing issues to worry about in a dynamic IP/MPLS mechanism, carrier-class connectivity, centralised NMS-based provisioning for hassle-free lightning-speed provisioning, and pre-provisioned paths and services that minimise re-convergence issues when network failures occur.

E2E SLA (service level agreement) monitoring will verify the Quality of Service in the network and allow the service providers‘ customers to verify online that their SLA is met, and there will be E2E CIR and EIR guarantees for every data and CES service via a best-in-class network management system, as well as a IEEE1588-2008 “Diamond Lane” for Timing over Packet to ensure minimal delay through the network.