Much of the vendor messaging at Mobile World Congress 2013 was dedicated to the concept of the intelligent network. Just to pick a couple, Cisco came with "The intelligent Network" strapline while Tekelec launched its "ThinkingNetworks" tagline and Tektronix took the wraps of its TIP (Telecoms Intelligence Provider) strategy. You could add many more.
At its core, network intelligence is about enabling agile service providers to be able to differentiate through the network, but more importantly in the services they offer over them.
However, network intelligence can just be too all-encompassing as a term, and become devoid of meaning. Where is network intelligence to be added, how is it to be managed, what do we want intelligent networks to be intelligent about, and what for (to what end)?
At the top level, the strategy rather than the tactics, if you like, is the move to SDN, or service provider SDN (a distinction made by Ericsson to demarcate from the data centre environment and to emphasise the cross-network nature of its SDN vision). The absolute bottom level driver here is cost and opex reduction, by centralising control of the network - so that the actual hardware-based infrastructure becomes as cheap and "dumb" as possible, where possible/viable. On top of that, by distributing resources more intelligently, the idea is that operators are able to enrich services with information about the network, services, or applications.
So then you might move into how you can best enable the intelligent network within this overall SDN end-game. Here are some of the topics that lend themselves to this network intelligence discussion:
RAN developments: This might include something like active antennas that can adapt to real time conditions. Or the pooling of baseband resources. Or the addition of increased processing at the edge. It could also refer to the integration of WiFi-cellular technologies and the control of traffic flows across the two dependent on what the operator knows about cell conditions, the user and the application. Finally, there is the introduction of small cells, a new layer of the RAN integrated with the core.
Signalling, policy and DPI: If you are going to have a centralised control function, allied to this is the need to be able to inform that control layer with real time information about the network itself, and its users, so that the decisions the controller/network orchestrator is making are based on sound information and are then effectively implemented based on scaleable, flexible control (signalling) architectures.
Monitoring and assurance: Information is key in building intelligence, and there's a huge amount of information sitting ready to tap, across interfaces, on devices, in the network nodes themselves. There's a bunch of companies whose traditional role of using that information for test, assurance and network management can be leveraged into something that feeds the intelligence layer.
Virtualisation: the core enabling technology of the flexible network that can allocate resources based on demand, policy, network conditions and all the rest of it. Which functions in the network lend themselves best to virtualisation? It promises to have a big impact on the platform providers.
Optimisation: The requirement for networks to self-organise and self-optimise in the light of what they know about the user, traffic and cell conditions is a key part of the network intelligence movement.
Over today and tomorrow, The Mobile Network will be publishing interviews and stories from Mobile World Congress that review this theme of network intelligence. We will let the industry itself tell the story. Here is Part One. Part Two, with a new cast list, comes tomorrow.
Building the Intelligent Network
Andy Ory, CEO, Acme Packet: Something interesting happens when networks move from TDM to IP. When a network moves to IP, and I mean IP origination, transport and termination, when that happens there's a fundamental separation of the delivery of services from the services themselves. Service providers ask themselves a very fundamental question and that is, do I want to get my network involved in offering services or not?
I think that almost all the service providers want to be involved in the latter, that's how they make money, build their communities, differentiate their connectivity. If they want to do that then they need to have service architectures that allow their IP networks to be involved in delivering these kinds of services; and they've never done that before.
I think what's interesting now is how are these networks going to be built, are they going to be built from NEPs that make the gear, from the IT side that can vertically integrate the services and the billing, OA&M and application enablement? So it's interesting to look at what the future looks like.
See Acme Packet's Announcements From Mobile World Congress
Jan Hagelund, Head of IP & Broadband, Ericsson: Elasticity is a popular word in the data centre industry but you can have end-to-end elasticity across the whole network in terms of computational power and processing.
The cloud execution environment is the execution environment across the whole network on which apps run in virtualised way. We are evolving into the Ericsson Cloud System, an open applications execution environment that we execute across the whole network. The most important thing is that from an end user perspective you have the opportunity of distributing applications to where they benefit from being in the network, enhancing the end user experience with lower latency, increased computational power. So what we are doing is puttin the same virtualised execution environment across different types of platforms, everything ranging from the data centre to data centre-orientated specialised hardware platforms to a traditional telecom-grade blade server execution environment, to our Smart Services Router, which is our app-orientated router platform. Then further out into the network we are also including the access side.
The SDN is how you control the underlying network.
How do you make sure the network supports the evolution so that network assets are presented to apps in a straightforward way. In a traditional architecture this is fairly complex, because you have lots of network nodes with their own capabilities and services. What the SDN does is centralise the control of the network and thus present capabilities for exposing services to those apps in a much more straightforward way, and essentially make the network programmable. So the combination is a very powerful combination.
I think SDN will have a role in different parts but maybe three main parts. One, obviously in the datacentre to simplify switching and the expansion of datacentre capability. The other place is on the IP edge where you have all these user plane services; by adding SDN you can configure those user plane services in a more straightforward way. And I think you can also conceive SDN further out in the metro and backhaul network, and that's more to do with complexity. You have many different nodes in metro backhaul so that by centralising the control and service definition of those nodes, you get quicker time to market for new service.
Ericsson demonstrates Service Provider SDN vision at Mobile World Congress
Thorsten Robrecht, Head of Portfolio Management, Mobile Broadband, Nokia Siemens Networks: For the past two to three years we have had the Liquid Network umbrella, which is about bringing intelligence into the network to steer resources and get the best out of the resources of a given network. It started with Liquid Radio, with things like active antenna, beam forming, putting resources on the air interface to use it in the best way. We are now shipping active anntenas in volume. Then there was baseband pooling, one benefit of which is now demonstrated by the way we can deliver FDD-TDD simultaneously in one SingleRAN base station using spectrum in same way for same device. Then there was our Liquid Core, where we are continuously following our virtualisation strategy with our products, meaning we are leaning towards all our core elements on the same commercial ATCA type of hardware.
And then on top of all, what we are launching new this year is linking the IT environment even closer into the mobile infrastructure environment, bringing an IT blade into the base station with Liquid Applications.
Nokia Siemens Networks at Mobile World Congress 2013
Andy Ory, Acme Packet: IP networks are all about smart stateful edges and a dumb core. The high ground, the real complexity is smart, stateful signalled management of secure media at every network edge. And what's interesting is our SMX is really all about a much simpler core. So our contention is you can be fully IMS 3GPP compliant, which is important on the RAN side because people need to have interoperability of devices and it's important when carriers want to hand traffic of, but we think that IMS ultimately is also going to look a lot like the internet. So the way Google and others deliver services with load balancers and so on is the way IMS needs to be architected inside these IP networks as well. So our IMS core with SMX and our SBCs at the edge allow you to be much more web-centric in how you deploy services, where you locate services and how you accommodate new innovation and how you dynamically expand capacity.
It's really all about the edge, but that's why we started getting involved in the core as well. The core is a requirement to make it simple.
The value that Acme Packet has in SDNs is this belief that networks need to be evolved to secure and manage stateful media at network edges at very low cost, and at the core there are web-like fully IMS compliant control plane technologies.
Jan Haglund, Ericsson: The way networks are configured today - where a function is placed is not necessarily where a function will be placed tomorrow. I don't think one size fits all in the paradigm. What we are doing with the Cloud Systems and Service Provider SDNs is to introduce more freedom of distribution, and I think we will see different network configurations: we don't need to mind as much as we did in the past. If it's the same execution environment it's really cloud management of where to put functionality.
For example, CDNs today are very centralised but with that increase in traffic it will make sense to have CDNs further out in the network because we then have a critical mass of enough people in the same neighbourhood viewing content.
Thorsten Robrecht, NSN: The key technology is inside the mobile network: we've got a lot of information which today is just using for steering the network but is not being made available for any type of user topic: how many users are where, going in which speed and direction, what type of plan do they have, the data they are consuming, the different type of devices they have? All this is available but we are using it today just to steer the network and control the resources. Imagine making it available to add value to other applications. Our Liquid Applications concept and RACS (Radio Applications Cloud Server) does this.
Signalling, DPI and Policy:
Jason Emery, Assistant Vice President, Product Management, Tekelec: We think Diameter routing is one of the first things that gets virtualised. It provides the signalling infrastructure that allocates and orchestrates resources, decides which function needs resources at which times. DPI today is unintelligent in terms of deciding what data flows, what traffic goes to the DPI. Networks will need to be more intelligent into what flows go to the DPI, so that policy-defined thinking networks can scale.
With SDNs, you have to virtualise in a way that allows the orchestrator itself to be orchestrated. There has to be a way to manage the growth and that means you must manage the control itself. People make the same mistake over and over again with signalling and control. If you are changing the network architecture, you need to manage control and signalling associated with that.
Tekelec's Software Defined, Policy-driven thinking networks
Jan Haglund. Ericsson: If you take an edge router today, typically all the traffic passes through a predefined set of services; firewalling, packet inspection, throttling. As an operator you have to decide on one size that fits all for traffic. But for certain users and services you could have that traffic pass throughput different capabilities; that would save network resources because you are not passing through everything and you are also making it possible for more individual offerings for end users. So what it does is say that using policy you can configure the network on an individual per subscriber, per flow basis and set up whatever network capabilities you want for that flow. And then SDN is perfect because you are centralising that network control - that's what policy based service chaining is.
Andy Ory, Acme Packet: We see Diameter as a protocol for the delivery of data-based services over LTE networks. It is a control based technology, not subscriber orientated, but we care about it because it is an entree into the rest of the control plane of subscriber based services, which is really what a SDN does: subscriber based services, the delivery of VoLTE and other types of services over IP.
Ken Osowski, Director of Solutions Marketing, Procera Networks: DPI has gone from analytics and traffic management all the way to revenue generating services based on policy, associating traffic information with a user through northbound interfaces either to the PCRF or the OSSBSS systems, focussing on value based services that are subscriber orientated.
So for instance, Yoigo took OTT VoIP and designated a Skype user signature, and instead of allocating that usage to the overall data usage plan, it created another bucket of data allocated for usage of VoIP, charged to the user by the minute.
I think that for intelligence associated with SDNs we should ask: Where does the intelligence come to build the SDN network? It's going to come from DPI in a lot of cases, by understanding what's going on in the network. Network intelligence is over-used as it relates to SDN but it is very well understood as it relates to what we [DPI] do. People understanding the behaviours not just of the network but of the subscriber, and what they are doing with applications so services can be personalised, is key to the monetisation of any network going forward.
Procera Networks and Tilera Unleash 200Gbps Deep Packet Inspection Solution
TOMORROW, Part II of "Building the Intelligent Network" features:
SON, Assurance, RAN innovation, platform innovation, and a whole new cast list of speakers.