Left to right: Scott Stevens, vice president, technology at Juniper Networks; Glenn
Dasmalchi, Juniper’s global enterprise CTO; Alan Burkitt-Gray, editor of Global
Telecoms Business (chairing the debate); Nils Swart, product manager in
Juniper’s strategy and marketing organisation; David Noguer Bau, head of
service provider marketing EMEA for Juniper
      
Global Telecoms Business hosted a round table at Juniper Networks' New Network Technical Forum in Barcelona in October, at which Juniper executives discussed the issues that delegates from the world's telecoms operators had been reviewing at the event.
                                                       
Alan Burkitt-Gray: Scott, can you tell me what sort of trends and changes you’ve been noticing in the service provider market, not only from your background in Juniper but also from this conference.
Scott Stevens: This conference was interesting. We spent a lot of time talking about the changes that we’re seeing as service providers start to move from focusing on transport and delivery of connectivity, delivery of IP infrastructure, and start to figure out how to be more strategically relevant to their customers.
How do they think about the kinds of applications their customers consumer and how do they become part of that conversation? It becomes a much more interesting conversation about evolving the business model. The business model has bifurcated between the content that gets delivered to the customers and the network that provides that delivery. Those two worlds have ignored each other — since the beginning of IP.
This has worked historically because a lot of the revenue growth for service providers over the last decade has been based on customer acquisition, so the faster they add customers the faster the revenue grows. The growth curve has pretty much plateaued. That leads to a fundamental shift in the business model. The land grab is pretty much at an end. How do they continue to grow revenue?
When carriers were growing revenue fast because they were adding customers it was easier to overbuild the infrastructure. Until recently most applications that were accessing the network weren’t necessarily that demanding. With all the video services — over-the-top video especially and video conferencing, and unified communications, and all of the different sets of high bandwidth applications that are coming out — bandwidth growth is exceeding network growth for the first time in our history.
The business model is changing, which is putting pressure on the network which is getting more congested. Those applications that expect the network to be infinite aren’t going to get that. The business model of application providers that expect full access to whatever they need may not be as easy as it sometimes was in the past.
What if the network could tell the applications: here’s what my capabilities are, and the applications could ask for what they need from the network? Today the network is trying to guess and the apps are trying to guess. What if we changed that ecosystem? 
                               
In the past they were guessing?
Stevens: And adapting. If you’re streaming a video to your iPad and the network gets congested the video streaming site backs off and tries less resolution and more compression, and if that fails the screen freezes. How do we build the world where that stops happening, where we stop the screens from freezing?
If you look at what’s happening to cloud, you are talking about more centralised content. You get greater efficiency with large hosting centres. Cloud computing is going to put more stress on the network. We are starting to talk about interesting sets of APIs that can occur to allow these two worlds to create a feedback loop.
How do we present a set of formalised APIs to allow the interesting things that need to happen more effectively? There are some strategies that we’ve implemented, creating a set of SDKs or APIs within the network. We talk about what we can do to open up the Junos platform — we’re opening up the R&D of the world. 
                               
Who would be these third parties doing this?
Stevens: We’ve got examples of functions that are now able to be embedded within the router that have better scale and support for the kinds of things we have to do. We also see some service providers building features and functions directly with the SDK themselves to solve what may be a unique problem for them.
With these open APIs, there is the facility to create feedback loops, so something that happens in the network can talk to the orchestration layer, or the application can talk to the orchestration layer, and talk to the client, and affect how the network behaves. 
                               
Glenn, what is going to be the impact of the whole range of cloud services on the telco’s operations.
Glenn Dasmalchi: There’s going to be a very large impact. There’s a lot of opportunity to provide cloud services. When we talk to enterprise IT, it is about delivering complete applications and tying those to business value.
There are some interesting use cases, mostly a mixture of private cloud and a telco-supplied cloud, and there is networking within the data centre to provide those services, there is networking that connects the data centres together, and then finally there are users.
If you’re talking about employees, they are bringing technology that they are used to from their consumer lives — the consumerisation of IT, with mobile devices. They’re used to be able to use these services from wherever they happen to be, not just on a corporate campus but anywhere in the world.
We believe that new architectures are needed from the network. We’ve made investments in technology at Juniper, including one that is a way to rethink the data centre, to flatten it out into a single tier. The performance characteristics are must better. The security characteristics are also better: security changes in this cloud world because you have bits of applications that are distributed around and need to be protected.
Once you have these basic features in the network you have to link it in to the IT systems. How the network integrates into the system is the final piece of this overall puzzle. 
                               
There have been some sad stories about cloud in the past few months, some problems. Are people as confident about cloud within telcos and companies like Juniper as they were a year or so ago?
Dasmalchi: We’re in a world today where a lot of things are still being called cloud. What was called outsourcing or co-lo we’re calling cloud, because that’s the fashionable term. The headline will always be: “The cloud has gone down.” But that’s not the case.
It’s about people being comfortable with the new technology. There are some applications that will need to be run internally, because maybe they’re subject to some regulations, or maybe the data has to stay within the EU. But external cloud abilities are going to improve and it will be more appropriate to send more to the external cloud. There’s always going to be this blending, what we call the hybrid cloud. 
                               
Who’se leading this work? Is it people in telcos who have spotted market opportunities?
Dasmalchi: Some of the earlier moves were pioneered by the over-the-tops, but there is a limit to how far they can go in terms of SLA capability and security capability. Telcos own the network. They are intimately familiar with their customers and they have contractual relationships with their customers. And so they have a big opportunity to provide cloud services that meet an SLA, that provide compliance, that are based on a trusting relationship that they over-the-tops don’t have the opportunity to provide. 
                               
Nils, you and Juniper have been looking at network programmability, addressing some of the problems with cloud.
Nils Swart: If you look at any given network you have four players — end user, application, content that is desired by the user, and the network that connects everything up together. So far these four entities have lived in relative harmony with each other — when harmony is defined as they don’t really talk to each other.
Most of the service providers know everything about the network and everything about the end user — or most of what the end user ought to be receiving from a telecommunications service.
A true over-the-top player knows everything about the application and that is intimately tied to the content, and it knows from the periphery what the end user looks like, but it doesn’t know how to get to the end user nor does it know what the end user is consuming in addition to what it’s trying to offer.
Likewise if you’re only an application provider you have very little control over how your application is being delivered to the end user.
What we’re suggesting with network programmability is to cut down the wall between the application world and the end users by means of bringing applications, content and end users closer together. Traditionally what we’ve seen is network programmability equates with how to write software in a very network-centric way. What we’re saying is that network programmability should be almost exactly the opposite. We cannot ask applications to become network-savvy and that has mixed results.
On the network side, there’s a lot of spying on the traffic, and that is successful to a certain degree, but not really. There should be an information flow from applications into the network with a clear description of what they want. Being able to give that clear description is one of the challenges.
We’re asking application developers to have an interest in what the network can offer. One way we’re trying to get them to learn about this is to tell them what information is inherently there. 
                               
You’re making use of existing information that’s there in the system?
Swart: Absolutely. Information as and where regulatory compliance allows for that. To turn them from greedy applications to something that’s more pleasant.
For example, if the network is able to tell the application to hold off for a little bit before pushing a petabyte worth of data through, you can generate new business models. [For example the network can say]: you can do it now and it costs 100% [of the standard price], do it in two hours for 75% or do it whenever I signal it to you and then it will be half the cost. 
                               
Is the information available already or are you at Juniper having to put in bits of information through the networks so you can see what’s going on and what the demands are?
Swart: The information is already there. The billing profile is widely understood by one department in a large telco, but mostly not by other departments inside that same telco. The problem of applications and networks not really talking to each other can be found within an enterprise itself as well as within an operator itself — using the OSS/BSS to scrape the information where possible rather than the network giving out exactly what it wants.
We’re suggesting a new set of APIs should be made available to those orchestration layers. Then you can go to the next step of deciding how to place traffic on the network in a way that doesn’t require a networking PhD.
How do you make sure that only the traffic that is either paid for or selected gets moved on to that particular path? OpenFlow is a very interesting technology to go and select which traffic goes from a receiving port to a specific path. OpenFlow is an industry initiative to split control of the forwarding plane from the actual forwarding plane itself.
It’s going to be a multitude of APIs that will be made available so a thin orchestration layer can be built, upon which application developers can write their application. 
                               
You mentioned regulatory compliance. Is that something telcos and application developers have to be alert to? Are we getting into the area of network neutrality? How do you get round the worries that some regulators may have?
Swart: We’re in a very interesting time when it comes down to network neutrality. The ability to identify network traffic but also to give out the right types of interfaces will hopefully create a level playing field that is useful for customers of operators as well as OTTs that want to traverse the operators. Yes, regulatory compliance is always an issue and something we take a close look at. 
                               
David, you’ve been here at this conference with lots of service providers. What are they telling you?
David Noguer Bau: They are in the middle of an identity crisis and they need to redefine their business. Some have already tried, and they’ve defined themselves as a cloud provider or a CDN provider. They are looking at ways of redefining their model. They still have their core business, connecting customers.
We need to define what is the value of a service provider. With intelligence about the customer, they can deliver services such as IPTV — they know the network, they know the set-top box, they control the environment, they know the bandwidth available. How can they open this up to third parties, such as over-the-top providers that are looking for ways to provide the content?
The user experience will be the next thing. Wherever there is an SLA, there is business for them. 
                               
Is there an opportunity where operators act as wholesale providers with local loop unbundlers? That sounds as though it could be another complexity between network providers, service providers, over-the-top providers and companies like Juniper?
Noguer Bau: This is another case with the European market. There are different models in the EU. It is splitting even more the value chain. In some countries content owners are merging more and more with the network space.
Getting network and content together doesn’t mean they have to be in the exact same company. They need to be able to interact through contractual agreements. What they get out of it is a better use of the technology.
We see lots of experiments. Many customers have signed up to the Junos SDK. Hopefully we will see it soon in commercial deployments. 
                               
Are we seeing collaboration between Juniper and other equipment providers? For all you would love it, you’re not going to get a service provider that has 100% Juniper equipment and they will have to work with other service providers in the real world as well.
Noguer Bau: The technology is there. Standardisation like OpenFlow are good examples of this. Juniper has experience of this when we joined the IPsphere initiative — getting SLAs across multiple service providers. Now it’s part of the TM Forum. We understand that it’s a mixed world: Juniper and multiple vendors.
Swart: Some vendors will be sooner to market than others. It’s safe to say that now is the time to make the network more relevant to applications and make the applications more relevant to the network. 
                               
Glenn, is there a return on investment for telcos in providing cloud services?
Dasmalchi: There is. Providers need to be careful about what services they are planning to offer. A lot of enterprises are looking for cloud services with enterprise-class SLAs and enterprise-class security. Telcos are in a great position to provide that relative to the OTTs.
They have to look at what goes into the cloud stack. A lot are based on virtual machines and hypervisors — that might not always be the right answer. You need to look at the cost and licensing models.
The networking equipment is very important for efficiency and cost reasons, but also in terms of its ability to support security isolation capabilities with multiple tenants in a cloud but also SLA requirements as well. There is a business there.
Stevens: The service providers are in a very powerful place to make an offer in cloud. Their data centres are behind the security blanket that is part of the enterprise’s network. An enterprise buys VPN connectivity from a service provider, so all the enterprise’s locations are on that VPN. The data centre in the service provider’s architecture that is extending the cloud reach into the data centre is on that same VPN. So you have a data centre that looks just like that enterprise’s data centre that is now a service provider offering. We can deliver SLAs on a per-flow basis on a specific session from a specific consumer to a specific data centre.
How does the network do exactly what is required to make sure an application is consumed in a way that’s useful. If you’re talking to a CIO, that’s really what they care about. The service provider has a powerful story that will answer the concerns of the CIO.
Programmability of the network will allow service providers to create offers that haven’t been made before. If you’re a service provider that moves early, you’re going to take market share and you will have a more valuable offer to your customers. 
                               
Are there people already moving early? Are there exciting things out there?
Stevens: Yes, I know of a few — they’re not public yet — that are based around this philosophy. These carriers are taking customers back that they’ve lost, that are coming back because the new offer is so much more valuable than what was available before — not a cheaper offer but a better offer.
Swart: A lot of those telcos are working together, similar to the way that IP peering was done years ago. Some of us are surprised that the telcos have moved so quickly to provide cloud federation on the back end — to make up for geographical reach, to produce a better offer than the OTTs.
Noguer Bau: We’re seeing service providers say their service is a cloud-ready network. Those telcos that are able to do this, they are really offering a cloud networking service. The combined offer of a cloud service or application together with the network is unbeatable. This is where they will build differentiation.
Stevens: It will be interesting to see which application developers will move first, along with the network providers. That will require some new skills. Over-the-top providers have a lot to offer and can benefit as well — when we get to open interfaces over-the-top providers can express what they desire from service providers, and the service providers can communicate back to the OTT: please don’t send so much traffic via one particular location, but via a different location.
Swart: An OTT can cease to be an OTT, and become a joint value proposition with the network provider. The ecosystem changes, where they are both making an offer. That’s what we’re seeing happen. That’s a pretty powerful place that will churn the business model. It becomes a collaboration. If you put the two together it grows the pie, it becomes a more valuable set of offers that can be exploited by the network provider and the application provider. They’re going to have some pretty big competitive advantages. GTB 
                               
Further reading from Global Telecoms Business:
Experience delivery networking: the network ahead 26 Sep 2011
NSN and Juniper for Fastweb IP backbone 29 Jun 2011
Juniper Networks offers $1bn in debt 01 Mar 2011