CARRIER ETHERNET BUSINESS BRIEFING 2014: Into the cloud era
14 February 2014 | Guy Matthews
Effort is now required to bring Carrier Ethernet into line with a cloud-driven world. Guy Matthews looks at the Cloud Ethernet Forum, and considers its likely next steps.
When the term “Carrier Ethernet” was coined nearly 10 years ago at a Metro Ethernet Forum (MEF) gathering in Hawaii, the concept of cloud computing as we know it had yet to be formulated.
If anybody present at that meeting had described a data centre where tens of thousands of virtual machines resided together, providing services to locations hundreds, or even thousands of miles away, they would have been greeted, at best as a visionary, but more likely as a fantasist.
It its first incarnation, Carrier Ethernet was simply not tooled to support such a vision. Now its backers know they must move with speed to adapt the standard for a market where cloud-based data and applications are becoming the norm.
May 2013 saw the birth of the Cloud Ethernet Forum (CEF), an MEF-affiliated body whose mission is to help cloud service providers, carriers and enterprises to support cloud services more easily, quickly and economically.
Its eclectic initial membership – Alcatel-Lucent, Avaya, Juniper Networks, Spirent Communications, Equinix, HP, PCCW Global, Tata Communications and Verizon – tells a story in itself. It is a list clearly hand-picked to reflect the diverse nature of today’s communications and computing universe, spanning as it does manufacturing and testing, cloud service provision, data centres and carrier services.
The forum’s work will not be complete until it has developed Ethernet to enable millions of virtual servers and storage devices to perform resiliently and efficiently across regional and global networks. It must take Ethernet’s basic VLAN capabilities and scale them dramatically, consider the issue of network performance at Layer 2 level and build in the high level of resilience that the cloud requires to be viable.
The forum has already announced its intentions to disentangle the sometimes fraught relationship between Layer 2 and Layer 3 protocols, reduce the stifling complexity of much networking equipment and cut back on the number of network management layers cloud operators have to deal with. No pressure, then.
At least Ethernet is already the fundamental connection of choice into and out of a typical data centre, and also the glue that binds all the servers together within the facility, says James Walker, VP of managed network services for Tata Communications and president of the CEF.
“The job of the CEF is to make sure that Ethernet is as appropriate as possible for this job, and that it continues to grow with the cloud, particularly when global geographic coverage is needed,” he says. “We’ve got to map it on to the needs of cloud service providers.”
Walker describes the needs of these providers as both “subtle” and “profound”.
“Cloud demands a lot more of a network than regular packet switching does,” he points out. “Its applications are very much affected by latency. And where you are trafficking in personal information, you need to know when a packet of data has gone over jurisdictional borders, perhaps exceeding parameters set for it. If the cloud provider can’t meet certain parameters, then they might even have to shut down the service.”
He says that none of the CEF’s goals would be feasible without the wide cross-section of its membership.
“You’ve got names on there you don’t normally see on connectivity forums, like HP, which is a vendor but also a major professional services organisation too,” he says. “Now we’re looking to extend that even more widely.”
He says the forum has already passed its first milestone, with the publishing of a white paper that sets out the organisation’s stall and explains some key concepts – in itself a useful dry run for the sort of collaboration that will need to take place if anything meaningful comes of the initiative.
Important feedback from the white paper has already made its way onto the CEF agenda in critical areas like security, claims Walker. He says the forum is being careful to garnish feedback from as wide a geographic catchment as possible.
“We’ve had a regional meeting of the CEF in Asia-Pacific, which shows we intend to involve different parts of the world in the process,” he explains. “We’re currently putting together technical committees to take up some of the issues raised so far.”
The forum’s hot early pace should play well with carriers and cloud service providers, eager for concrete results in short order. But history tells us that the work of such bodies is seldom to be rushed to any sort of final, comprehensive conclusion. The MEF itself was established 11 years ago and seems nowhere near done with the streamlining of network interconnection. Walker rejects the notion that the CEF should have started its work far sooner.
“The cloud has grown faster than anyone expected,” he insists. “A lot of the work that was being done before the CEF was set up is complementary to cloud issues, while some of the questions we’re now tackling have only emerged recently.”
The biggest challenge
Perhaps the CEF’s biggest headache is to reconcile the speed of evolution of cloud services with the much-more-gradual changes in direction that can be managed by the juggernaut of global networking.
“Cloud connectivity is important, but in the early phases,” as Matthias Hain, director of data services with COLT puts it. “But cloud services themselves are by no means in the early phase of adoption. Connectivity is now a crucial issue for large enterprises, meaning there is a place for operators to be a one stop shop. There’s a need to put the correct capacity underneath the IT.”
The matter is one of culture as much as anything. In the IT world, the principle of consumption according to need is the norm. There is plenty of discussion of moving the carrier business model more into line with this approach, but talk is always easier than action.
“It won’t be overnight, but a matter of steps on the journey,” says Hain. “The steps might be pay-by-use, then bandwidth-on-demand, going on to customers directly manipulating what they want. Carriers need to go steadily here, to make sure they come out of it in profit. They certainly can’t stay as they are.”
Phil Tilley is a senior director with vendor Alcatel-Lucent, responsible for strategy development for its IP routing and transport portfolio. He is excited by the progress he is seeing in data centres, where the job of orchestrating, automating and virtualising is in full flow.
“Cloud service providers are ahead of where we are with the network,” he admits. “There’s some way to go in network automation. But there’s plenty of evidence that the delivery of cloud over Carrier Ethernet is happening now, where it wasn’t a couple of years ago.”
The next challenge is to take cloud service providers, consumer-facing communications service providers and backbone operators and join those dots together to forge a three-way connection.
“That way you can be sure that you can extract the right feature set from the network, so that the application knows what to do,” says Tilley.
The most pressing task for the CEF, he feels, is to bring together virtualisation and data centre people, networking equipment vendors and service providers, and guide them towards a common mindset.
“The IT and the network providers are working to optimise Ethernet, so that the network delivery side is at the same level of automation as the virtualisation of the data,” he says. “We’ve got the players together now. But we’re not going to fix all the problems immediately. It’s a continuous effort. You can use Carrier Ethernet for things like cloud today, but you don’t have the automation you’ll have in two or three years – hopefully.”
The next step
Tilley says that a major step forward has been made with the specifications of Carrier Ethernet 2.0.
“CE 2.0 makes Ethernet more robust, and that’s step one,” he believes. “We’ve seen a big uptake of that by vendors and now by carriers. It’s a good foundation. At Alcatel-Lucent we can see both the opportunity and the challenge of bringing all the elements together in the next phase.”
Geoff Bennett, director of solutions and technology at vendor Infinera, believes the bar must be set as high as possible by the equipment-manufacturing community, to ensure nobody bangs their head against early limits.
“We’ve got to be aiming high on transport, to provide support for cloud, not to mention other areas like M2M, video, mobility,” he says. “Transport infrastructure needs to respond to the growth in Ethernet’s capacity potential. Ethernet is the only client service that’s working at 100G. Does the underlying transport match that? We’re shipping 500G equipment in the belief that it’s only by allowing that sort of oxygen that you make client services possible. Can we build Terabit Ethernet? No, not yet. The electronics don’t support that.”
Vendors looking to the future must also provide the technology to extend virtual machines beyond Layer 2, says David Noguer Bau, service provider solutions marketing manager for Juniper Networks.
“We’re doing this by configuring MPLS to use GRE tunnels,” he says. “As for connections from customer to cloud, I’m not seeing a lot of E-lines, unless you’re talking about a private cloud. A Layer 3 connection is more normal. What the user expects is flexibility, and flexible bandwidth in particular. This can make Ethernet the choice in other senses than basic E-line.”
There are, of course, other ways of visualising the cloud-to-network reconciliation challenge than those presented, sometimes as orthodoxies, by the CEF and MEF.
Steve Best, managing director, strategy, products and regulation at BT, says that Ethernet is central to the services the operator provides, including its support for cloud providers, but not necessarily according to a strict MEF interpretation. He sees the work of the Cloud Ethernet Forum as potentially valuable to a startup service provider establishing themselves in the cloud world – less so for a veteran incumbent like BT, with its long history of Ethernet connectivity.
“Ethernet is crucial to all the service areas we are going into,” he says. “We need to make sure we are providing a set of capabilities that the customer can pick and choose from, and we need to supply a very good Ethernet product to market at a very competitive price. I’m glad to say that we haven’t lost a single order for not having a more formal MEF relationship than we do.”
Another non-MEF-standardised approach to cloud support is Masergy Marketplace, from Texas-based operator Masergy. Scott Bender, the company’s marketing communications manager, nonetheless says it is the perfect exchange-like forum for cloud service providers wanting an Ethernet-over-MPLS network.
“It is for people who want to connect their global fabric to key cloud providers,” he says. “We’ve got all that on-net now, with real-time visibility and control – not MEF-standard though.”
Matthew Finnie, CTO of European operator Interoute, is concerned that the global Ethernet community, for all its rhetoric, takes too inward-looking a view of the wider connectivity market.
“The flaw with networking services is that all too often they exist in a bubble,” he says. “They make an assumption that everything that happened in the past will just repeat itself. But the idea that Ethernet will always exist as a market on its own doesn’t ring true.”
The convergence of computing and networking leaves Carrier Ethernet looking like a thinning layer in a fast-growing sandwich, believes Finnie.
“It’s less and less relevant to talk about Ethernet as if it were something that matters on its own,” he complains. “Ethernet is not going to disappear, but we really need to see it for what it is.”
But one way or another, CEF-certified or not, Ethernet looks like the only choice out there if cloud services are to break through to the next level. It is the only solution that takes any kind of account of the next wave of data growth that is likely to swamp the cloud, the only one to offer efficiencies in provisioning and true scalability.
But it is only fair to conclude that the carrier industry is still at the stage of defining what the next generation of challenges are, without an all-encompassing solution necessarily being in sight.