Inter-carrier networks are here to stay

The challenges and opportunities this brings to wholesale service providers.’.

MEF members including leading wholesale Service Providers, cloud and datacenter providers, equipment vendors and peers from a range of telecom organizations discuss, debate and share views about the challenges and opportunities from the ever increasingly complex inter-carrier networking telco environment.

Mobile, wireline, cloud and data center service providers rely increasingly on third party access vendors (AVs) to fill gaps and increase footprint. Coverage is ensured, but often with deployment delays, compromised performance and quality – because off-net circuits across multiple carrier networks are a challenge.

How can telecom services achieve the same operational simplicity, dynamic service performance and real-time operational state that can be found in the cloud today? How can the reliance on off-net circuits stop being a burden? What is needed to support critical inter-carrier requirements, such as audit and inventory, ordering, fault isolation and SLA management? How can the operator accelerate delivery, reduce OPEX and boost service quality and agility? What role does SDN and NFV play in all this
 

Michael Howard launched the discussion with a brief outline of the current and anticipated market.

Firstly that the migration from TDM to Ethernet, though continuing apace, still has a long way to go. There is a lot of legacy stuff still out there.

Secondly: mobility – mobility everywhere. Customers want a service, but they want it to be available on the move. That means service providers are asked to connect customers with Ethernet anywhere, on-net and off-net. That in turn puts the emphasis on e-Access, further emphasised by the move to cloud services.

Thirdly, customers don’t just want the connection, they want services that work – and that is shifting the focus to SLAs and beyond to application performance and quality of experience.

Meanwhile the demand for Ethernet services is growing and carriers are responding with more and more new types of service with increasing demand for CE2.0 standards in RFPs. A March MEF survey showed the growing demand for CE2.0 e-Access certification: in the past 12 months, about 20% of RFPs required it, but in the following 12 months the figure was expected to rise to 65%.

In his first diagram Howard pointed out that CE equipment sales were reaching maturity, while services continue to soar. There is not a direct link between the two figures because CE equipment covers all the carriers’ internal infrastructure needs, only part of that being directly linked to service provision. The infrastructure may be in place, and the service ROI continues.

In the second diagram we see an example of increased spending on wholesale Ethernet services. Of the panellists present: Andrew MacFadzen estimated that Orange had around 600 partners, including 170 access providers across 120 countries worldwide; and James Walker estimated that Tata had over 270 last mile providers worldwide and they resell service from around 690 other carriers.

So Howard ended by asking the team, given that global context: what is needed to support inter-carrier requirements? An important MEF role has been to create some kind of standardisation to make it easier for the wholesale market, and that is what e-Access is about. But the technical issue is only the beginning: there is also audit, inventory, ordering, fault isolation, SLA management and other issues to be aligned when one carrier connects to another. We must address both the technical and the business aspects.

Speaking for the MEF, Nan Chen suggested three major areas to be addressed:
• Technical interconnect, about how to get the bits flowing from one network to the other. MEF 26.1 is their contribution to that, so far
• Service interconnectaddressing the two carriers’ different services, offerings, SLAs etc. This is where e-Access is so important – it has made it much easier, but there is still work needed on that.
• Process interconnect – this is the next big challenge and it is going to take a long time to make this connection seamless between carriers. This is the reason the MEF launched its Service Operation Committee (SOC) a year ago and addressing those problems.

To put these challenges in practical terms, Andrew McFadzen pointed out that Orange has around 170 access providers and, when the company first started using them it might take around nine months from choosing a partner to actually launching a really robust connected service.

He continued: “Replicate that across 170 partners, and you realise fairly quickly that it’s not scalable. So, we created a ‘network factory’ to slim down those processes – greatly aided by CE2.0 because everybody had a different terminology and definitions. Standardisation has dramatically reduced that initial time to engage with a new partner”.

So you now have a working partner, but that is just the start. If a customer asks for Ethernet access to an office in Manaus, they have to first ask their partner in Brazil for a feasibility study: can they do it? Next: how much will it cost? And then follows the ordering etc. “So, we are working on how to condense all of that down. Maybe havingsome link into their tools to ‘window shop’ and see where and what services are already available. Not rocket science, but stuff that hasn’t existed in the past and will help us activate services much quicker and get revenue faster”.

Howard responded with a quote from a carrier in CE’s early days: “When you buy an E1 or a T1 you know what you’ve got, at what speed. You connect it and you know what you’ve got”. Ethernet is wonderfully rich with variables, all of the factors that you could ever want, but that means way too much choice. In recent years, the MEF has narrowed those choices down, but there is still a rich set of options.

However, Mark Showalter of Infinera, speaking with experience of installing T1s, pointed out that it was not quite as simple as folks like to imagine: “It took years of standardisation to make that happen.”

Walker liked being able to have virtual links inside an environment controlled by a third party operator to: “run our version of the service over the top of somebody else’s infrastructure. So that abstracts away from some of the direct problems of aligning all our various operational processes or measurement processes or whatever with that far end carrier. You claim the ability to heap all of your tools into mine, but then your tools were written by somebody I’ve never had to interface with before, or they don’t have our level of detail in their system…We try to step over that problem by having visibility end-to-end of the circuit with either our own NID or a virtualised NID inside their infrastructure. Insight into their network is really important.”

Christopher Cullan from Infovista brought up the question of managing SLAs. Infovista did a global survey of Ethernet providers last year, asked to rate SLA reporting on a scale of 1 to 10 criticality in business terms - eg to customer acquisition and retention – and 9 was most frequently cited. There was little doubt about the need, but the problem lay more in deciding the right metrics, how to measure them and then make them easy for buyers to understand. This is part of their work with the MEF SOC.

According to McFadzen, however, Orange’s enterprise customers no longer put so much store on SLAs: “It’s gone down the priority list.”Typical SLAs to a retail customer cover site availability, packet loss, and very technical things, but customers are saying: “That’s nice, but really I want application performance more than the availability of the site”.

This is where CE2.0’s classes of service come in, according to Nan Chen. The MEF mapped 20 application types to class of service, to give providers general guidance in MEF 23.1 for, say, three classes of services to suit certain application types. Bob Mandeville added that it took a year and a half of lab work to measure these applications and translate them into the five metrics t in MEF 23.1.

The trouble is that the applications were categories like real time versus download video or voice, where real world applications are more like Amazon, or Salesforce. Walker pointed out “it is getting more and more like currency trading: one customer was moving servers as close as possible to Salesforce to cut latency to below 10ms, but Salesforce transaction times can take between 600 and 1400 ms – so they simply cannot deliver everything the customer wants.”

CE 2.0 has really raised the bar on performance expectations, according to Mandeville. For example: “We know that application traffic is bursting, so the MEF addressed this by developing the concept, of committed burst size. Not just bandwidth – we’re allowing application traffic to be bursting within that bandwidth.” The testing methodology and implementation took almost two years to develop and it took a commensurate time for vendors to develop the switches and routers capable of meeting this very high spec. “So, I think the MEF’s certification and testing is playing an important role in getting networks ready to support applications’ challenging new demands.

That led to the issue of application-based routing, and the role of SDN and NFV. The simplest way to look at SDN, said Nan Chen was to separate the control plane from the data plane – that is mostly levels zero to three. Then NFV is really about layers four to seven: “Ideally every single function in all of the network elements will be virtualised, based on the VMs’ capability”.

And yet: “It is going to be so difficult. Take inventory as an example: it used to be ports, cards, and connectivity – very simple. For Cullan the better example was fault isolation in a moving target: “Where are these things now? Or, where were they when that issue happened, so I can try and figure out and make sure it doesn’t happen again? I think that dynamic nature makes things far more complex from a management standpoint”.

According to Howard, carriers around the world insist that net revenue is the number one driver, because of the agility to quickly put up new services or take them down or modify them if they don’t work – such agility to bring revenues is the biggest driver. “But do carriers really need all that agility?” asked Walker. In the datacentre, fair enough, but to achieve that in a carrier network with SDN would be extremely computationally intensive:“passing enormous amounts of information off switches and routers into some sort of controller. The controller gets much larger, and closer to the switches it controls, so I need a huge backbone network to connect my controller to the edge! I absolutely see that there are benefits – very strongly when you talk about cloud environments and complex datacentres. But, it’s a complexity of interconnection. It’s not a complexity of a core nature.”

With that, it was time for Howard to bring the discussion to a close.