Metro Cloud – a growing opportunity for high speed data centre interconnect

This Industry Viewpoint was authored by the MEF.  

In the light of the MEF’s recent Third Network announcement, a round table discussion on the Metro Cloud opportunity arising from this announcement was held on 15th October in Düsseldorf, Germany.

This article summarises the main points in the discussion, which included the following participants.

Chaired by:

• Rick Talbot: Principal Analyst Optical Infrastructure, Current Analysis

Panellists:

• James Walker: Advisory Director, MEF Board, Founder & President CloudEthernet Forum, Vice President, Managed Network Services, Tata Communications
• Chris Purdy: Chief Technology Officer, CENX
• Chris Liou: Vice President of Network Strategy, Infinera
• James Armstrong: Executive Vice President, Spirent Communications
• Andrew McFadzen: Head of Global Marketing, Network Services, Orange Business Services; Chairman, MEF

Talbot began by outlining his view that the opportunity was for operators to provide more Ethernet services, while currently most datacentre connections are set up via full-time private lines. Often the customer simply leases the fibre, then installs and manages its own link – so the provider misses the chance to deliver a bespoke, value added service. The difficulty of cracking this opportunity has been that setting up a Carrier Ethernet (CE 2.0) service is still a manual process that can take weeks to align to customer requirements. If, however the Third Network makes it possible to deliver that capacity on demand – or at least to connect the datacentres in a fraction of the time that it would take to build on a leased line – then the service provider (SP) would have a very attractive business offering.

Speaking for CENX, Purdy noted that large, well-established datacentre operators had the resources to build their own high speed links, as well as providing dedicated connection to major customers – thereby bypassing the SPs. But other operators rely on their SPs and so do not want to compete with them – and that puts them at a disadvantage, because each time they have to wait around 90 days for the SP to establish the link to the customer.

SPs that could deliver connection on demand would pick up a lot more business – although Walker pointed out that the real attraction would be when the customer needed a fairly complex link between several datacentres, whereas a simple link between two datacentres would be a tougher sell. According to Walker, SPs are divided into two main groups: the dark fibre providers that focus on delivering a transport layer with little in the way of value added services, versus the enterprise-oriented providers that offer more additional services such as firewalls, traffic optimisation etc.

According to McFadzen, there is actually less talk now about datacentres, the main question becoming “do you want your services hosted in the cloud?” Orange have a hybrid network approach that can either provide connectivity into a customer’s datacentre or access and services to some 25 Orange datacentres around the world – or else access to third party datacentres.

There was general agreement that this cloud migration was greatly increasing the demand for connectivity. Speaking for Infinera, Liou added the growing importance of network function virtualization (NFV) with customers linking datacentres across a metro area to create a single virtual datacentre with dynamic bandwidth connection to the Internet or to cloud providers. The trend towards greater agility suggests a great business opportunity: “sell them 40 gig Ethernet for weekend back ups as a service” rather than having lengthy turn up times and the resulting pressure to keep the circuit in service.

To clarify the distinction between these “metro cloud” datacentres and the giant ones currently making the news, Purdy explained that the latter were not your average idea of a service provider: “First, they’ve got the money to build these warehouse-sized datacentres. A lot of them are using those datacentres more as a backbone for their operations.” He quoted some statistics that suggested that every kilobyte Facebook request generates a thousand times as much network traffic as the original request – new traffic generated in their datacentres because of the way the service was developed. And the average Google search request actually traverses a crazy 1,500 miles, creating new levels of traffic between their private datacentres. So that sort of user is not so interested in bandwidth on demand as simply owning their own massive pipes.

For a normal service provider, the bandwidth needed between datacentres is growing at so fast a pace that CENX sees a real market opportunity for Ethernet over optical transport geared specifically to the metro datacentre market – high density, stackable systems offering a more flexible service offering.

The conversation turned to the SPs for comment, and Walker confirmed the two different markets: the web scale providers buying 100G and 400G wavelengths, and the Enterprise market. He explained the 1,500 miles of a Google request, saying that each request goes to three separate datacentres and the first to get an answer is the one that replies – so web scale providers’ networks are designed to distribute, and resiliency is less of an issue, whereas enterprise systems are designed to survive: “[the web scale provider] can build datacentres comparatively cheaply, like a warehouse where land is cheap and there are tax incentives, with not much air conditioning, air filtration or massive fire suppression systems – because it’s only one of three that is going to respond to the query. Enterprises have to have a datacentre that will to survive typhoons, fire, power outages, all that kind of thing”. So providers to the enterprise market – like BT or Tata – have to spend far more on their datacentres and could not compete in the web provider market (where the provider would rather build its own network anyway). “If a Google query doesn’t get cleared, nobody dies usually. If somebody can’t phone the ambulance services in France, then the chances are that somebody will die”.

This also implies that SPs like Tata cannot risk experimental technologies: “SDN is not yet appropriate for the types of things that major carriers like Orange and Tata and BT and so on do because it is designed to operate within a very closely densely tied together datacentre estate with one customer operating over infrastructure you probably own anyway.” McFadzen also explained that, apart from Orange’s own giant datacentres in France, Spain and Poland, they are linking many rather smaller ones in order to provide proximity and low latency to the customer base, and these will also be linked together, or to third party datacentres, in order to provide disaster recovery.

As a carrier, Orange is responsible for connecting its own datacentres, whereas a datacentre operator would be more likely to outsource the links to an SP, as Armstrong confirmed: “If they start competing with the service providers, then the rest of the service providers could pull out of their datacentre and their datacentre becomes disconnected and loses value.” He went on to point out how Third Network agility would practically eliminate this dilemma: if some SP could deliver the right connectivity without all the delay, most datacentre operators would choose to outsource it, because they would rather concentrate on their core datacentre business.

McFadzen, however, sounded a note of caution: agility is very much flavour of the year, but do customers really want to chop and change as much as the visionaries think they do? While there are without doubt certain sectors where agility is key to competitive advantage, there is a great deal of solid business where the customer might prefer a fixed rate service to “pay as you use”: “On the first of the month you’ll be wondering how large your bill is going to be. As a CIO, Head of IT of whatever, do you want that? Or, do you want to understand and have predictability of a monthly recurring fee because you’re paying for a fixed level of service and at least you know how much this year it is going to cost”.

How much agility is enough? Purdy suggested that in general: “The value of going from 90 days to 30 days is huge. To go from 30 days to one day is still valuable. But to go from one day to one minute?” That really is a niche requirement, and yet it would still be an advantage for the SP to be able to promise that service if it was needed.

One other point was raised by Walker: providing connectivity in the way under discussion is very much about getting the optimum return on existing infrastructure, and being able to respond quickly to demand for services on it. If the connection requires installation of new infrastructure, then the delay is more about the time needed for physical installation and, above all, for getting permission to do it: “In the German market, for example, the ability to lay new fibre is very limited because city councils don’t like having their roads dug up. In fact, especially in Germany it’s almost impossible to dig new fibre. So, anybody who has fibre and ducting would rather hang onto it than sell it off wholesale”. So you tend to find a small number of providers who focus on owning dark fibre, while the customers go to people like AT&T or Verizon for actual business services.

The above summarises the main points in the discussion. The full transcript is now available at LINK.

If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!

Categories: Cloud Computing · Ethernet · Industry Viewpoint · Metro fiber