In Mobile Europe's February 2007 issue, Alun Lewis looked at the development of 3GPP standard LTE. now he looks at its counterpoint – System Architecture Evolution. The shape of things to come, he says, is flat, uses packets and needs policies…
After the post-recession doldrums of a few years ago, it's fast becoming clear that many late ‘90s visions of the evolving ‘network of networks' are finally starting to become reality. While the future then was going to be based on such faded and largely forgotten technologies such as ISDN and ATM – as well as early versions of cellular and fixed radio access – many of the core principles remain the same, despite the huge impact of IP, DSL, WiMax and 3/4G on network architectures. That original vision was based on the idea of a single, access technology-agnostic, core network that spanned both the fixed and mobile world, long before terms like fixed-mobile convergence had even been thought of.
For the last year or so, most of the headlines in the industry's trade press have focused on the radio aspects of this new network world – defined by the 3GPP under its Long Term Evolution (LTE) initiative, planned as part of Release 8. Running slightly behind in the standardisation stakes – around eighteen months or so according to some industry experts – are complementary and equally important developments in the shape of the core network of the future currently known as either System Architecture Evolution (SAE) or Evolved Packet Core (EPC).
While the main focus of LTE is on, as ever, increasing the bandwidth and performance of the radio links, SAE has additional concerns to attend to – with much bigger and broader strategic issues to be addressed that will decide the commercial shape of our industry for years to come.
The prime driver behind SAE is the overwhelming need to reduce latency in the network – essentially the delays in accessing and using applications and services. If the industry is make the mobile internet experience a real part of everyday life then there's no point subjecting users to the wireless equivalent of the World Wide Wait that so frustrated users of the fixed internet back in the 1990s. While the telecoms industry engages itself in constant debate around the issue of ‘when is real time really real time' – and the answers do tend to largely depend on whether the speaker has a background in IT or communications engineering – the harsh reality is that much of the industry's expected future revenues do depend on services that require a near-instant response from an increasingly distributed set of back office, application and content assets and resources.
There's no point offering VoIP, streaming media, gaming, PoC, rich IM, Mobile TV, Web 2.0 or multimedia content if the user experience is hindered by slow transaction times. We've all become sufficiently aware of consumer dynamics to know that even just one or two bad experiences with a new application will put the customer off and inhibit the experimentation mindset that's so helped drive the fixed internet.
In parallel, there's also the continuing need to break down the old system silos that have grown up over the years and eliminate old and increasingly costly dedicated infrastructures, systems and processes. For a network of networks to become a reality, there need to be ways, for example, to ensure that policy, authentication, security and QoS parameters can be applied consistently – even while the customer is moving between different access technologies such as between 2G to 3G and WiMax and between different devices and network operators .
So how is SAE going to achieve this nirvana of instant gratification, seamless services and lower capital and operating costs for the network owner ?
Essentially SAE is all about creating a flat, all-IP core network that can handle all the new demands – highlighted above – that will be made on service providers in the coming years. As ever, the devil will be in the final detail but it is important to recognise the close relationships between developments underway around the all-SIP world of IMS and, in a wider context, the work underway on TISPAN through the ITU and ETSI with more IP-specific input also coming from the IETF. In terms of 3GPP standardisation activity, significant dates are only a few months away, with decisions being agreed on the core network and on supporting protocols planned for December 2007 and March 2008 respectively.
For Paul Steinberg, a Motorola Fellow, these initiatives are final proof that the old network model is ‘finally unravelling'. As he explains, "Mobile service providers are obviously looking to eliminate costs by simplifying the network and improve the customer experience by reducing latency and increasing bandwidth. On top of that, they also now have the challenges of fixed-mobile convergence and the emergence of WiMAX encouraging them to move towards an all-IP core capable of controlling while managing far more complex interactions between an ever-more diverse set of devices and applications. The fixed-mobile aspect seems to particularly relevant at the moment and here the links between SAE and TISPAN are especially relevant."
This is a perspective echoed by HP's EMEA CTO, Brian Levy who comments that, "TISPAN is ultimately a lot more interesting in this wider context than IMS, especially as it goes a lot further into the peer-to-peer realm where more and more traffic is going to be coming from. Ultimately, it's becoming clear that the service paradigm at the end-points will be much more like the world of the internet than traditional cellular communications."
Certainly the shift towards an all-IP environment is going to bring its own set of new issues to confront with the total elimination of circuit-switched infrastructure. While the picture still seems unclear about the role of much-heralded IPv6 in this new world it is certainly true to say that many quality, policy and security factors have yet to be fully resolved – as do questions about how SAE will fit into current or near-future OSSs and BSSs.
As HP's Levy adds, "The world we're facing is going to need a combination of a smart network with smart devices. The potential role of IPv6 is strong here given the imminent explosion in the number of always-on, permanently connected devices that will require their own addresses and that will at times be controlled by other devices – such as will be occurring in interactions between say mobile devices and IPTV in the home. Not only do these types of transaction need to be handled securely and reliably with policies and SLAs reaching down to various devices, but we also have the secondary issues of how to consistently enforce QoS policies across networks owned by different operators. The IPSphere Forum is already actively examining some of these issues to link business models more closely with the underlying IP infrastructure."
According to Preston Gilmer, vice president of product marketing at OSS vendor Sigma Systems, "When it comes to policy and QoS, OSSs will have to interface with a policy server or Policy Decision Function (PDF) to provide the relevant service profile and CPE IP or MAC address to ensure that the appropriate QoS policy level can be applied to the network or related application server. In a SIP environment there can actually be multiple SIP sessions that apply simultaneously to a single SIP CPE device. In this case, a policy can be set at a network level and each application or service consumed by a user can have an associated policy applied. This is an example of ‘service level' orchestration coordinated and managed through an OSS – even though all the relevant network and application and service level policies are actually managed by the policy server of PDF."
In talking about differentiated policies, issues to do with charging and billing are never far behind and once again we run across the complexities of cross-network/cross operator relationships. Interconnect billing has always been a notoriously difficult area – and one frequently prone to revenue leakage through error and, at times, blatant fraud. Fergus O'Reilly, director of product marketing at charging and billing specialist Highdeal believes that, "While a lot of important work has already been done in the 3GPP charging groups, such as TSG SA WG5, this is now being rejigged for Release 8 through work being done jointly by ATIS (the US Alliance for Telecommunications Industry Solutions) and 3GPP, 3GPP2 and the ITU, leading to common standards for charging across the NGN space.
He also highlights the continued need for recognition of how customers are going to interact with the increasingly dynamic pricing models demanded by a very open network world. "Telecommunications providers are already creating complex bundles of voice, data and entertainment services," he adds. "As we open up networks and move between different access technologies, customers are going to rely on Electronic Programme Guides being able to display prices relevant to that customer, at that particular time, on that particular device."
Recognising the incredibly granular nature of NGN services and applications – essentially extending the already highly personalised nature of mobile communications – creates correspondingly high levels of complexity in policy issues which are themselves further complicated by the different heritages of the industry sub-sectors involved. Farshid Mohammadi, vice president product line management at policy management company Bridgewater Systems, sees, "A recognition at Release 8 by the 3GPP community that policy has become far more important and is going to need to be applied across multiple domains. The standards may have evolved separately but they are now converging together – and much of the focus is going to have to be on subscriber-centric policy management and finding ways to propagate these across different networks and devices, while simultaneously mapping customer requests to the overall assets available at that time in the network. The application needs to know what the end points of the network are and what limitations these will impose on service quality."
In this context – and also with wider concerns about identity and privacy in mind – SAE and related developments place far greater emphasis on the individual's identity than ever before. For Shane O'Flynn, vice president of managed services at event and transaction management specialists Openet we're going to need to integrate a number of functions and features such as security, QoS and charging policies and aggregate these around the customer's individual identity in a similar way to that carried out by the role of the Policy Charging Role Function (PCRF) in current IMS architectures. "Only in this way," he comments, "are we going to make a wide range of services ‘sticky' enough, above and beyond the basic connectivity, especially given that we'll be working in an environment where the traditional boundaries of pre and post-paid, on and off-net will be breaking down."
Security across a heterogenous network environment will apparently be dealt with in an IPv6 world through Ipsec, but that in turn could latency issues as Othmar Kyas, director strategic marketing network diagnostics at test and measurement vendor Tektronix suggests, "On the basis of early trials that we have seen, implementing Ipsec can be quite a challenge with the use of public and private keys and ciphering adding to the device processing payload as well as raising configuration questions for the customer. That said, Ipsec is very secure."
So, the gameplan for the next few years looks reasonably well agreed with the usual range of intricate and interconnected issues to be resolved. The question we might as well start asking ourselves now is what comes next?