White Paper

The Multi-Access IP Edge: Key To Delivering FMC Prepared by Patrick Donegan Senior Analyst, Heavy Reading

www.heavyreading.com

TABLE OF CONTENTS I.

THE MASS MARKET OF REAL-TIME IP APPLICATIONS................................. 3

II.

DEVICE PROLIFERATION ACROSS IP EDGE NETWORKS............................. 4

III.

NETWORK CONVERGENCE AT THE IP EDGE ................................................. 6

IV.

CONVERGENCE AND THE MULTI ACCESS EDGE ........................................ 11

V.

REMOVING BARRIERS TO THE MULTI-ACCESS EDGE ............................... 13

VI.

SUMMARY AND CONCLUSIONS ..................................................................... 16

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

2

I.

The Mass Market Of Real-Time IP Applications

Adoption of real-time multi-media IP applications, and convergence of those applications across multiple access networks, are the two defining trends in the delivery of new communications services. In wireline networks, there are now more than 50 million VOIP users world-wide and more than 12 million IPTV subscribers. In mobile networks, real time IP services might be in their infancy today but they are bound to see large scale adoption. This is all but inevitable from 3GPP R7, due to be commercialized in 2009, when W-CDMA operators will be able to deliver mass market cellular VOIP services with significantly more efficient use of available radio spectrum than they can achieve today delivering voice over dedicated bearers. The adoption of IP-based services started in fixed and cellular networks more or less independently of one another. As a result IP sessions that support interoperability between different access networks and device types are still limited in number today, and are still the preserve of a small number of users that happen to be served by a small elite of leading edge operators. By leveraging the ubiquity of IP, it is the goal of carriers to add value to these real-time multi-media sessions by delivering Fixed Mobile Convergence services which are differentiated from discrete fixed or discrete mobile IP services by two main characteristics: the ability to access them on any terminal, across multiple different fixed and mobile access networks (such as, for example, videotelephony sessions between fixed and mobile terminals); hand-off mid-session across any availlable fixed or mobile access networks with little or no perceptible impact on the user experience. That service providers are intent on executing on fixed mobile convergence services is clear from the findings of a Heavy Reading carrier survey conducted in December 2007. Figure 1: Carriers See FMC Services As a Major Revenue Opportunity

Source: Heavy Reading Carrier Survey, December 2007 n=138 To date, the telecom industry has made significant progress in serving up some of the tools that telcos need to deliver on this promise. From a consumer perspective, the most obvious example of this is in the form of converged devices. Take a look at any laptop today and it has ports that

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

3

can support connectivity into an enterprise LAN, consumer DSL, WiFi, various iterations of 2G and 3G cellular, 56 kbit/s dial up and, pretty soon, WiMAX. In practice, while the ability to hand off from one network to another with little or no impact on the user experience is relatively easy to achieve in some inter-network hand-off scenarios, it is a lot more challenging in others. Nevertheless, the fundamental multi-access capability in devices exists and will increasingly be built into other devices as well, like set top boxes that support multiple wireless air interfaces. The industry has also made great strides in enabling convergence at the network level: •

Converged standards in the core: many FMC services have, of course, been delivered on proprietary network platforms. But significant progress has also been made in specifying, and in some cases implementing, the Internet Multimedia Subsystem (IMS) standard. This is designed to provide a standards-based core network environment for supporting generic Session Initiation Protocol (SIP)-based sessions which are access-agnostic.

•

Converged transport networks: In transport networks, integrated service providers with both wireline and wireless access networks are converging transport resources to lower their cost of operations. Specifically in the mobile access network, operators are investing in additional backhaul capacity to support increases in cellular data traffic enabled by High Speed Packet Access (HSPA) and 1X EV-DO upgrades to the air interface.

II.

Device Proliferation Across IP Edge Networks

Whilst great strides have been made in upgrading the IP core and transport networks for the requirements of discrete and converged SIP-based services, comparatively little attention has been paid to the IP Edge. In a telecom carrier service paradigm, it's at the IP edge of the network – between the access network and the core – where IP sessions are aggregated, authenticated, secured, and where policy is applied. A brief overview of the current network architecture at the IP Edge, as well as the requirements of a profitable mass market in converged IP services, suggest a growing case for convergence of network elements within this strategical part of the network.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

4

As shown in figure 2 below, today each discrete access network has its own dedicated edge gateway element for carrying out those functions – the Edge Router, BRAS, WAG, ASN Gateway and GGSN/PDSN. Figure 2: Discrete IP Edge Gateways In Today's Network

Source: Heavy Reading In the early days of delivering basic IP data services across both fixed and mobile access networks, these edge gateway devices were deployed more or less in isolation, with perhaps the exception of some basic security appliances. In wireline networks that was fine ten years ago when most of the data traffic was delay tolerant applications like email and web browsing. In mobile networks that was fine five years ago when most of the data traffic was delay tolerant WAP and text messages. When Internet innovators initiated mass market take-up of VOIP and other real time services over IP across wireline networks, it triggered a whole a chain of industry responses. Whilst the investment of huge resources in standardizing IMS is the most celebrated outcome of this new era, another key impact was the generation of a requirement for a variety of new session management capabilities to compliment the primary edge gateway. These new session management capabilities were needed to re-introduce an overlay of conventional carrier-class control capabilities to deliver real-time multi-media IP applications with the right quality of experience to the end user. In fixed network environments, the two primary product types that have had to be deployed alongside the Edge Router and BRAS at the edge have been security gateways and session border controllers: •

Security Gateway: These have been deployed to authenticate users and secure both signaling and media streams by encasing them in encrypted tunnels. This sort of security is aimed primarily at protecting subscriber traffic and making sure that subscribers get only the level of service they requested.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

5

•

Session Border Controllers (SBCs): Though primarily used over the past couple of years to solve IP-to-IP connectivity issues at the network-to-network interface (NNI) between peering partners, SBCs have also performed rudimentary functions, such as firewall and NAT traversal, at the access portion of the network. As carriers expand their networks to support hundreds of thousands of VOIP users, however, a new breed of SBCs is starting to be groomed for the edge of the network. This is on the grounds that management profiles need to be applied to real-time sessions as soon as possible once the session enters the carrier's network in the interests both of maximizing the user experience and making best of use of the relatively limited available bandwidth at the edge of the network.

As the number and variety of real-time IP sessions in the network increases, so the depth and breadth of session management functionality is having to increase in step. The same trend is in place in mobile networks where new functionality has been deployed at the edge to support an array of session management functionality such as content billing, content monitoring, traffic optimization, content filtering and content caching. In addition to these, new mobility management functionality has also had to be added to the mobile network to support hand-off of IP sessions between cells when the mobile user is on the move. •

III.

Mobility Management: In a discrete cellular network, the mobility management function is responsible for tracking the user's movement between cells. In the GSM/W-CDMA standards, mobility management for IP sessions is supported in the Serving GPRS Support Node (SGSN) as well as in the GGSN. In CDMA it is supported in the Packet Data Support Node (PDSN). In WiMAX and WiFi, mobility management is also supported in the ASN gateway and Wireless Access Gateway (WAG) respectively.

Network Convergence At The IP Edge

The flourishing of supporting session management functionality distributed across multiple devices around the edge gateway is already becoming a cause for concern as regards the management of any one access network. In the case of carriers that support many – or all – of these different access networks, this proliferation risks generating even greater cost and performance barriers to the optimal delivery of fixed mobile convergence services. The different opportunities and risks associated with distributing functionality across multiple types of network element, or alternatively, consolidating that functionality into fewer nodes, have often featured at the heart of network planning deliberations. The merits of each approach tend to be re-evaluated whenever CTO organizations seek to prepare their network for a new phase in service delivery. The arguments are well known. At a very high level, consolidating functionality creates the scope for substantially reducing OPEX by reducing the number of nodes in the network, reducing the number of different device-types that need to be managed in the network, and reducing the number of transport links between the devices. On the other hand, centralized functionality increases the carriers' dependence on one vendor, and weakens the business case for the carrier to shoparound and buy-in 'best of breed' point solutions as and when they are launched on the market. At some point or another in their careers, many CTOs have presided over extremes of both consolidated and distributed models. Many have designed and overseen a change in architecture from one model to the other, and many can lay claim to have designed shifts that have taken different networks (sometimes even the same network) in both of the different directions. Both models – and the almost limitless variations on each – have always flourished in the telecom market. However, in an era of growing volumes of IP sessions being admitted and managed across multiple access networks, and of integrated fixed and mobile carriers looking to FMC services, there are clearly a number of factors that are increasingly favourable to consolidation at the IP Edge.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

6

The case for consolidation is being driven by five fundamental networking considerations that arise in the quest to deliver the new generation of discrete fixed, discrete mobile and converged fixed-mobile IP applications.

3.1

Higher Scalability In Edge Network Elements.

As demonstrated by the VOIP and IPTV subscriber figures cited in the introduction, carriers have barely scratched the surface of the potential of real-time IP services. No one much doubts that the wave in real-time IP services is going to get a lot bigger. The only issue is one of timing. As shown in Figure 3 below, consumer broadband connections continue to grow. Over the year to June 2007 the number of broadband subscribers in the OECD increased 24% from 178 million to 221 million subscribers. This growth increased broadband penetration rates in the OECD from 15.1 in June 2006 to 18.8 subscriptions per 100 inhabitants one year later. Growth in mobile broadband is behind the curve in terms of subscriber numbers but there are already more than 100 million CDMA 1X EV-DO subscribers worldwide today capable of supporting downloads at up to 3.1 Mbit/s. The much more widely adopted W-CDMA standard is also seeing strong ramp up in HSPA subscribers with more than 15 million HSPA devices in use supporting up to 7.2Mbit/s. This wave will mean a lot more than just more connections. Just as important as the continued ramp in subscriber connections will be the increase in the number of IP end-points per connection. Increasingly, a consumer DSL connection no longer serves a single PC in the home using email and web browsing. It might support three PCs using a variety of different real and non real-time voice and data applications. It might also support IPTV as well as perhaps dedicated WiFi phones or dual mode WiFi/cellular phones.

Figure 3: Broadband Penetration In G7 Countries Broadband penetration, historic, G7 countries 25 Canada United Kingdo m

20

France Japan

15

Germany

10

United States OECD

5

Italy

0 2001

2002-Q2

2002

2003-Q2

2003

2004-Q2

2004

2005-Q2

2005

2006-Q2

2006

2007-Q2

Source : OECD

Source: OECD This trend will increase in the wireline environment and is also being echoed in the wireless market. Here carriers increasingly project service scenarios in which EV-DO, HSDPA or WiMAX radios are deployed increasingly widely across a range of consumer electronics devices, household goods, cars and others so that one subscriber will have multiple devices optimized for different applications. And there are already examples of this beginning to happen.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

7

The standards roadmaps also make provision for step changes in the throughput that can be supported in the access network. This is particularly true of the wireless environment where the introduction of HSPA enables a W-CDMA device to support up to 7.2 Mbit/s in the downlink over the air today compared with 384Kbit/s two years ago. In the cellular network, the impact of HSPA and EV-DO is causing shockwaves in some operators with many reporting data traffic up three or four fold within three to four months of commercial launch. According to another Heavy Reading survey of 67 mobile operators in December 2007, 45% of cellular operators believe that data will account for more than 50% of their traffic by 2010. One major European mobile operator has told Heavy Reading that having seen data traffic increase between three and four-fold in 2007 due to HSPA, it expects a further eight-fold increase in data tariff during 2008. The same survey revealed that a typical urban cell site had 9 Mbit/s of backhaul capacity (equating to around five E1s or 6 T1s). Survey respondents predicted that this requirement will increase three-fold to an average 25 Mbit/s of backhaul capacity at a typical urban cell site by 2010. 3GPP's roadmap also sees HSDPA's downlink capacity expanding to 14.4 Mbit/s, and on to 40 Mbit/s in the 2009 timeframe. The next generation of 3GPP technology, UTRAN Long Term Evolution (UTRAN LTE), has a roadmap which extends to as much as 300 Mbit/s in the downlink and more than 100 Mbit/s in the uplink, depending on the assumptions used. In consumer broadband dedicated access bandwidth per customer is now more than 10Mbit/s on the ADSL2+ links that are now becoming ubiquitous in telco broadband networks, and most telcos are now planning a transitioning to either fiber to the curb (FTTC) delivering 50Mbit/s or more per customer, or fiber to the premise FTTP) delivering 100Bit/s or more per customer. FTTP. In a PON, “burst” speeds extend up to many hundreds of megabits, and next-generation PONs that deliver 10Gbit/s or more per split (typically to 32 customers) are on the drawing board. In point to point or active Ethernet deployments, 1Gbit/s speeds are already being offered in a few territories. The trends highlighted above point to strong growth in broadband connections across all access networks; more real-time sessions per connection; an increasing diversity of real-time sessions; and a lot more throughput going through each of the access networks. It's therefore clear that the edge network architecture depicted in figure two needs to be scaled upwards to keep in step with these developments. And if it scaled on the basis of just adding more of the same devices per access network, the challenges already identified with aligning this architecture to the new opportunity in real-time IP services will simply escalate.

3.2

Network Latency

Carriers throughout the world have identified reducing network latency as a critical requirement for delivering converged real-time IP services across multiple access networks that customers will pay for. The Next Generation Mobile Networks (NGMN) group of carriers comprising Sprint Nextel, NTT DoCoMo, China Mobile, Vodafone, Orange, T-Mobile, and KPN Mobile has an objective of reducing the end to end system round trip time within 3GPP and WiMAX defined mobile networks at 20 milliseconds. In a converged service environment, a session that has to undergo fresh authentication and the application of multiple session management profiles across multiple platforms each time the session is handed off from one access network to another is likely to be more vulnerable to missing the carrier's latency targets – and the user's expectations of a converged service experience – than one where admission and session management are handled centrally and by the same network element as the session is handed off from one access network to the next. (Requires validation)

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

8

3.3

The Highly Dynamic Character Of The Internet Applications Environment

. The very dynamic character of the Internet applications environment, featuring very short life cycles and allowing very limited carrier control, should also be a key consideration for network planners. The speed at which some applications become tremendously popular and then fade away is increasingly difficult for carriers to predict. It's early days in terms of users being able to access the same services from multiple different devices but predicting not only which specific applications will see take up – but also which specific devices users are most likely to want to access them from - is also extremely challenging. As an example, many carriers have been surprised to see the unexpectedly high proportion of mobile TV usage that takes place within the home, where people might have been expected to watch the program on their TV or even their PC. What this points to is that the IP edge network architecture needs to be one which is highly flexible and which enables the carrier to respond dynamically to rapid, and often short-lived, changes in the applications that people use and the access network that they access them from. The more flexible the IP edge network architecture is, the less is the requirement for the carrier to try and predict ahead of time exactly which way the market is going to go and harden the network's design accordingly. The greater the flexibility, the lesser the investment risk associated with having to "guess right" in an increasingly unpredictable environment.

3.4

Greater Commonality In Edge Network Requirements

In the early days of the IP services paradigm, the requirements for managing subscriber traffic at the edge of the network were completely different in the case of an edge router, a BRAS, and a GGSN or PDSN. In the wireline environment, the design requirements of an edge router and a BRAS were differentiated by the requirement for an edge router to support minimum SLAs for business subscribers whereas they weren't required for a consumer DSL service. That distinction has since been blurred due in part to the greater number of IP end points served by a single DSL connection, in part to rising consumer expectations, but primarily as a result of the rise of IPTV. These factors have transformed the QoS and throughput requirements of a traditional BRAS so that they now align more closely with those of an edge router. Historically, the distinction between the needs of edge gateway devices for fixed and mobile networks has been even more pronounced than that between different fixed access gateways. In complete contrast to their wireline counterparts, mobile edge gateways have traditionally been designed for hundreds of thousands of subscribers and small amounts of throughput per user. Equally, and again in complete contrast to their wireline counterparts, the first generations of mobile edge gateways have been designed with a lot of built-in control plane intelligence – to support mobility as well as real-time transactions such as billing for prepay subscribers – and a little bit of data plane capacity. Hence at the platform level a typical GGSN today has the majority of slots in the chassis dedicated to the control plane and the minority to the data plane whereas the inverse is true of a BRAS or edge router. As shown in figure 4, Mobile gateway designs are now needing to re-align in favour of data plane capacity to support 1X EV-DO, HSPA and LTE. And wireline gateway designs are now needing to realign in favour of control plane capacity to support better QoS across more simultaneous sessions. Hence the gap between the design requirements for supporting the two basic types of fixed and mobile access network at the IP Edge is narrowing.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

9

Figure 4: Evolving Edge Gateway Design Requirements For Fixed And Mobile Networks

Source: Heavy Reading The same is potentially true, though not yet, of session management features. There are some areas where the different standards bodies directing the requirements for the edge network have mandated the same protocols. For example, there's quite a lot of commonality in the session management protocols that standards bodies have selected for WiMAX, cable, DSL and FTTx. All four use MD5 and HTTP digest for authentication. All four use QOS shaping, BW policing, and CAC for traffic shaping as well as Signaling-ALG, Media-FW, and IP-FW for firewalling. They also all use IPsec for secure access, although in the case of cable, IPSec is only optional. However, a lot of differences still remain in terms of session management protocols in use across these networks. Moreover there remain stark differences in the session management protocols in use between the above four standards and 3GPP. While the specific session management protocols are often different across different access networks today, the fact remains that to a large extent many of these differing protocols are nevertheless performing very similar tasks to one another. In some cases, the reason one has been chosen over another doesn't reflect different requirements. Rather it simply reflects the historical legacy that standards developed for different access networks were developed in isolation from one another. Therefore it also remains true that the support of multiple different flavours of session management functionality at the edge– and the associated requirement in many cases for specific devices for specific networks – represents duplication of functionality across access networks and additional complexity and cost arising from the requirement to support so many different protocols.

3.5

"Mobile Phones" Are Becoming Consumer Electronics Devices

Many models of smartphones increasingly resemble mainstream consumer electronics devices. At the same time the consumer electronics industry is evolving into one designed to support more and more networking between devices. Examples of this are the efforts of the UnPlug'n Play (UPnP) Forum and Digital Living Network Alliance (DLNA) to enable the exchange of media/content within the consumer's residential network.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

10

As cellular devices increasingly become part of this "community" of networked consumer electronics devices within the residential network, this will introduce challenges in the operator's network because these devices will need to be permanently connected to both the cellular network and to other consumer electronics devices in the residential network via other types of wired and wireless connectivity. Mobility in and out of the residential network will also need to be supported.

IV.

Convergence and the Multi Access Edge

Figure 5 depicts the variety of consolidation options for the Multi-Access IP Edge, many of which operators have already started to invest in. The five options can be thought of as either autonomous, stand-alone options, or alternatively as sequential, interdependent options leading to option 5. Approaches 1, 2 and 4 are already in commercial deployment by leading carriers while approaches 3 and 5 are not yet commercially available. Figure 5: Multiple Approaches To The Multi-Access Edge

Source: Heavy Reading

1. Convergence of Session Management Devices Per Access Network. This is already happening today. Here the carrier integrates the session management requirements for a single access network into a consolidated platform rather than deploy multiple devices. In wireline networks, the boundaries between security gateways and session border controllers is blurring now as vendors from both backgrounds begin to compliment their product roadmaps to support the other functionality with a view to providing a single platform to support both feature sets. In the mobile network, some GSM and CDMA operators have already started deploying platforms parallel to the GGSN or PDSN that bundle content billing, content monitoring, optimization, WAP gateway, packet inspection, content filtering and content caching features in a single platform rather than support them as separate point solutions. This approach doesn't reduce the variety of edge gateways required per access type. However it does provide some consolidation on the session management side.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

11

2. Integration of Session Management features Into The Edge Gateway Per Access Network. This is already happening today. Here the carrier integrates a number of session management functions directly onto the edge gateway in hardware and software, thereby eliminating altogether the need for a dedicated box to support those functions. Vendors are already trialing this capability in wireline networks, for example with the integration of deep packet inspection features into both BRAS and edge router products. In cellular networks, the capability has been embedded in some GGSN products dating back to the 2004-2005 timeframe, and has been seeing significant traction in the last 18 months with operators staring to load capabilities such as deep packet inspection onto the GGSN rather than deploy a dedicated box. Optimus, the Portuguese carrier, is one among many GSM/W-CDMA operators world-wide that has loaded a variety of session management functionality onto its GGSN rather than deploy separate point solutions. Integration of Session Management features Into The Edge Gateway Per Access Network and the previous Convergence of Session Management Devices approaches differ in two main ways. The first is that the Convergence of Session Management Devices approach still leaves in place a dedicated session management node running parallel to the edge gateway in the network whereas the integration approach creates the opportunity to do away with one or more dedicated session management nodes altogether. The second is that integration of session management features in the edge gateway provides an opportunity to reduce network latency. 3. A Converged Session Management Platform for any network, fixed or wireless. This approach is not yet in commercial deployment. The opportunity here consists of leveraging the convergence of consumer and business requirements across fixed and wireless access networks with the transition to an IP applications environment. This approach could be delivered according to one of several distributed or converged models. A common session management solution could be designed for fixed networks, mobile networks, or both. It could be deployed as a discrete session management node in parallel to the edge gateway or else integrated into the edge gateway. Irrespective of the particular model chosen, convergence of session management would enable a large scale consolidation of functions which are currently replicated across multiple nodes serving multiple access networks. 4. A Partially Converged Edge Gateway supporting two or three specific access networks. Again, this is already happening in both fixed and wireless networks today. Here the operator makes a break with single-access edge gateways and converges all their fixed access traffic or all their mobile traffic onto a single access gateway. Leading carriers such as Bell South and Chunghwa Telcom have already started supporting both their corporate business subscribers and their consumer DSL subscribers from the same converged edge gateway device, commonly called a Multi Service Edge Router (MSER). In parallel, the first mobile operators – among the minority that support both CDMA and GSM/W-CDMA networks today – have started running their PDSN and GGSN functionality from a single chassis using a blade architecture. A model likely to see greater uptake is that of cellular operators combining their PDSN or GGSN with their ASN Gateway to enable them to support cellular and WiMAX subscribers from a single gateway. There are already converged cellular/WiMAX gateway products that are generally available on the market with more due in 2008. This ap-

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

12

proach allows the operator to take a major step forward in terms of reducing the duplication of gateway functionality in the network, although it still leaves the operator reliant on different edge gateway nodes to support its fixed and mobile subscribers. 5. A Fully Converged Gateway and Session Management Platform supporting any network, fixed or wireless This approach is not yet in commercial deployment. It enables the ultimate in edge network consolidation by converging the different edge access gateway functions onto a single gateway device capable of supporting any variety of wireless or fixed, consumer or business, connection from a single point in the network. A common session management solution capable of supporting all these different access types could also be supported, thereby enabling a dramatic simplification of the edge network. This approach would enable the transition from today's architecture supporting multiple gateways and multiple session management devices per access network to an edge network consisting of nodes that are all built on the same platform and all managing subscriber gateway and session management functionality in a centralized way.

V.

Removing Barriers To The Multi-Access Edge

The challenges of implementing approaches (1) and (2) and (4) of edge network consolidation are already being met in commercial deployment today. A number of important challenges still remain as regards approaches (3) the Converged Session Management Platform for any network, fixed or wireless and (5) the Fully Converged Gateway and Session Management Platform supporting any network, fixed or wireless There are three challenges in particular which need to be addressed: •

Overcoming organizational barriers within the carrier.

•

Aligning industry standards activity in a way that best supports fixed mobile convergence at the edge of the network;

•

Overcoming hardware and software design constraints which have traditionally made a viable multi-access edge device difficult to deliver to the market.

5.1

Organizational Challenges Within The Carriers

Organizational issues within the carrier can be a key barrier to convergence at the edge of the network. Fixed and wireless access networks have traditionally been planned, deployed and managed by different organizations, often with their own CTOs. As shown in figure 5 below, carrier surveyed by Heavy Reading at the end of 2007 identified organizational rivalries as the second biggest barrier to fixed mobile convergence. For as long as the access networks within the carrier continue to be managed autonomously, convergence at the IP edge will be confined to models 1-3 depicted in Figure 4. However, there is a clear trend among some of the world's leading integrated fixed and cellular carriers of converging their CTO organizations so as to reduce and eliminate the scope for that kind of internal rivalry. And it is these carriers that will be best placed to take advantage of the opportunities that a fully converged multi-access edge architecture has to offer.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

13

Figure 6: Carriers Identify Inter-Departmental Rivalry As A Key Barrier

Source: Heavy Reading Carrier Survey, December 2007, n=138

5.2

Aligning And Harmonizing Industry Standards

The achievement of the multi-access edge will require standards alignment across the multiple industry bodies developing the roadmaps for each of the access technologies. And it will require alignment across both the gateway domain and the session management domain. This is clearly not a straightforward task. But there are positive indicators in terms of recent industry precedent as well as some compelling reasons why many of the requirements will align more and more over time. IMS is the obvious example of a new emphasis on standards-alignment between industry standards bodies in pursuit of the Fixed Mobile Convergence service opportunity. IMS was initially defined by 3GPP, the GSM and W-CDMA standardization organization. Its work was then adopted by TISPAN as the basis for next generation IP services in wireline networks as well, and by Cable Labs on behalf of cable MSOs. This experience triggered a new network of inter-standards body liaison groups which are focused on driving commonality across standards wherever possible.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

14

Figure 7: 3GPP's Evolved Packet Core Supporting Non-3GPP Access

Source: Heavy Reading The convergence thinking behind IMS standardization for the core has already spilled over into standard-setting activity at the IP Edge, even though the output in terms of commercial product is a little further down the line. As shown in figure 7 above, 3GPP's roadmap for the Evolved Packet Core (EPC) in 3GPP R7 mandates that the GGSN will become an access-generic gateway capable of supporting non-3GPP access networks such as WiFi and WIMAX. Also, whereas GSM and W-CDMA are differentiated from WiMAX and CDMA 2000 in that the former use GPRS Tunneling Protocol (GTP) for mobility management of IP sessions and the latter use the Mobile IP (MIP) protocol, there is increasing alignment within these different standards organizations around using Proxy Mobile IP (PMIP) as the basis of a standard for inter-system hand-off between all the different wireless access technologies. This will be supported within a Mobility Management Entity (MME) shown as it could fit in a GSM/W-CDMA network architecture. The first EPC products allowing standards-based access to non 3GPP connections are expected in early 2009. The trends within telecom industry standardization in terms of closer alignment and harmonization between fixed and mobile access networks is clear..

5.3

Overcoming Software And Hardware Design Constraints

Developments in technology are enabling more flexible trade offs to be made between control and data plane processing. Advances in core technology and network equipment design are removing some of the design barriers to supporting fixed and mobile sessions in two main ways: •

Advances in core ASIC and FPGA technology. Trends in technology are making it possible to combine higher capacity, more intelligence and greater flexibility in a way that begins to break down some of the design barriers to combining fixed and mobile subscriber service management in the same multi access edge device. On the one hand, increasingly powerful network processors and FPGAs are enabling increased scalability. Alternatively, new generations of Packet Processing ASICs (PPAs) are a lot more flexible than traditional ASICs. This new generation of ASICs no longer requires hard coding. They have fully programmable micro code which doesn't require a hardware re-spin every time and which can be downloaded hitlessly. As packets transit through cards built with these ASICs, the individual chips have access to sections of the packets through low latency memory paths. The overall packet processing can be done on individual cores or pipelined through different cores for more

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

15

complex processing. The functionality that needs to be processed on a per packet basis (from an application perspective) would include functions such as subscriber identification, services within subscriber traffic and service specific treatment of traffic such as rate limiting, marking based on complex scheduling algorithms, as an example. Having multiple cores would enable the PPA to handle multiple streams simultaneously, enabling a scalable platform to be built. PPAs are already in commercial deployment by leading network equipment vendors. •

Leveraging the "third axis" in control plane scalability. There have traditionally been two axes in scaling control plane capacity in any network element: the number of card slots – typically 16 - in the chassis and the processing power associated with each of the two processors deployed on each card. It is this limitation which has traditionally dictated fundamentally different edge platforms for fixed and mobile. Leveraging new Packet Processing ASICs, it's now possible to scale the control plane according to a third axis – increasing the number of processors from two per card to eight per card. This is one approach that can enable the new multi access edge to support high scalability on both the control plane and the data plane.

VI.

Summary And Conclusions

While it may seem slow relative to the number of years it has been talked about, the communications services market is moving inexorably towards Fixed Mobile Convergence of IP-based services. The case for convergence at the edge of the network to align with convergence at the core and generate a mass market in profitable FMC services is overwhelming. The current trajectory of proliferation in the numbers and types of devices at the edge of the network is unsustainable in the long term. For carriers supporting multiple access networks it's not a question of whether to converge their network at the edge; it's a question of how far to converge it and in what timeframe. All five of the models outlined in this White Paper will see take-up, as well as a myriad of variations on each. As a general rule, the greater the subscriber numbers and the greater the number of access networks supported, the greater will be the incentive to converge the networks across both gateway and session management domains.

© HEAVY READING | APRIL 2008 | WHITE PAPER | THE MULTI-ACCESS IP EDGE: KEY TO DELIVERING FMC

16