hr epc platforms wp 0409

White Paper

LTE/SAE & the Evolved Packet Core: Technology Platforms & Implementation Choices

Prepared by

Gabriel Brown Senior Analyst, Heavy Reading

www.heavyreading.com

On behalf of Alcatel-Lucent

www.alcatel-lucent.com

April 2009

http://www.heavyreading.com/�

http://www.alcatel-lucent.com/�



© HEAVY READING | APRIL 2009 | WHITE PAPER | EVOLVED PACKET CORE TECHNOLOGY PLATFORMS 2

TABLE OF CONTENTS

I. INTRODUCTION & KEY FINDINGS ..................................................................... 3

1.1 Key Findings .......................................................................................................... 3 1.2 Report Scope & Structure ..................................................................................... 4

II. EPC MARKET REQUIREMENTS ......................................................................... 5

2.1 Outlook for Packet Data Services ......................................................................... 5 2.2 Functions of the Packet-Switched Core ................................................................ 6 2.3 Enabling Operator Business Objectives With EPC ............................................... 6

III. IMPLEMENTING THE EPC NETWORK ARCHITECTURE ................................. 7

3.1 Logical EPC Architecture ...................................................................................... 7 3.2 Integration of 2G/3G Packet-Switched Core With EPC ......................................... 8 3.3 New Platform Requirements ................................................................................. 9 3.4 Implementation Options ........................................................................................ 9 3.5 Alignment with IP/MPLS Core Networks ............................................................. 10

IV. PLATFORM ANALYSIS FOR EPC NODES ...................................................... 12

4.1 Economics of Platform Choices .......................................................................... 12 4.2 Vendor Strategies for EPC Platforms .................................................................. 12 4.3 EPC Bearer-Plane Requirements ....................................................................... 14 4.4 Toward the Access-Independent Gateway ......................................................... 15 4.5 Evolving Packet Gateway Requirements – 3G to EPC ....................................... 16

V. SAE GATEWAY PLATFORMS & OPERATOR CHOICES ................................ 18

5.1 Edge Router, ATCA, or Specialist Platform? ....................................................... 18 5.2 Platform Heritage & Economics .......................................................................... 18 5.3 Edge-Router Platforms ........................................................................................ 19 5.4 ATCA or Specialist Platforms .............................................................................. 19

VI. BACKGROUND TO THIS STUDY ...................................................................... 21

6.1 Original Research ................................................................................................ 21 6.2 About the Author ................................................................................................. 21 6.3 About Heavy Reading ......................................................................................... 21


I. Introduction & Key Findings Within five years, wireless access technologies will account for a greater number of global broad-band connections than DSL, cable modem, and fiber combined. To support the rapid increase in traffic, active users, and applications implied by this growth rate, mobile broadband operators need to rapidly transition to a flat, all-IP network with the Evolved Packet Core (EPC) at its center. The intent is to cost-effectively deliver superior application performance from the 3G/Long Term Evolution (LTE) radio access network (RAN) across the core network and provide end-to-end QoS in line with principles set out by the Next Generation Mobile Networks Alliance. The EPC should act as a "business machine" capable of assuring end-user quality of experience and im-plementing an operator's business rules as it seeks to maximize productivity from network assets. This report positions EPC in the wider market context and examines the major decision points faced by operators evaluating the implementation of packet core in LTE networks. With this as background, the second half of the paper focuses on the underlying technology requirements for the System Architecture Evolution (SAE) Gateway elements – the Serving Gateway (S-GW) and Packet Data Network Gateway (P-GW) – that interface between the RAN and external packet networks such as the Internet.

1.1 Key Findings The key findings of this report are as follows: EPC represents a fundamental shift in wireless network architectures, introducing chal-lenging new requirements to packet core equipment. With no circuit-switch domain specified for LTE, the packet core must support carrier-grade voice, differentiated services using end-to-end QoS, and vastly increased throughput per subscriber relative to 3G. Given the ten-year investment horizon, this is likely to mean that most deployed and currently available products will not fully meet operator requirements, which is driving discussion on which underlying hardware and software platforms are most appropriate for the various EPC deployment options. New hardware platforms will be required for EPC; many existing 3G products are not able to scale to support LTE. A majority of packet core vendors have outlined plans to introduce new hardware platforms for EPC, which reflects an inadequate legacy product base, and we expect all major suppliers to have revealed next-generation platform strategies within the next 12 months. While it is viable to support EPC applications on existing platforms and this will make sense in some cases, operators need visibility into platform roadmaps to reduce the risk of being left stranded with outmoded equipment. Platform choices will mirror the EPC architectural split, with edge routers used in the bearer plane for S-GW and P-GW, and blade server platforms used for the MME and PCRF in the control plane. This is a broad characterization, however, and not representative of the entire market. There's a subtle trade-off between platform types, with choices impacted by a vast range of operator- and vendor-specific circumstances. Despite this hedge to our opinion, we note that vendors with successful edge-router products have all opted to use this platform for SAE Gateways, and that those without have opted for an alternative type of platform. It is critical for operators to assess the resources suppliers are willing and able to invest in the platforms on which EPC products are based. The ability and incentive to invest in product development and provide long-term support is based on the revenue opportunity and/or strategic importance of a product portfolio. Mobile packet core is a relatively low unit-volume market, which has historically made it difficult for vendors to sustain and prioritize investment. For this reason, leveraging edge-router assets for SAE Gateways is attractive; however, the risk is that given the market size, vendors will not make EPC requirements a priority when developing or upgrading the router platform.


EPC will initially be deployed as an overlay in parallel to the 2G/3G core, but will ultimately become the converged core for all wireless access networks. This will minimize the risk of disruption to mainstream data services and allow operators to experiment with LTE ahead of mi-grating to a converged packet core for 2G/3G/4G. We estimate that within one to two years of commercializing and stabilizing the EPC, operators will want to start consolidating wireless access networks onto a common packet-switched core. It is at this point that software upgrades to deployed 3G equipment starts to become appealing. Throughput, latency, active user capacity, bearer setup times, and traffic analysis capability are the basic metrics used to determine SAE Gateway product requirements. Of these, raw throughput is the least challenging; but whereas in the past, packet processing capability could be sacrificed for greater control-plane capability, the move to EPC will also put great emphasis on latency "through the box" and wire-speed packet inspection. This creates a requirement for faster hardware and aligns SAE Gateways more closely with advanced wireline products. SAE Gateways will likely be collocated with 3G equipment in centralized locations initially, but could be distributed toward the edge of the network over time. Potential benefits of a distributed model include traffic optimization, improved redundancy, and local Internet break out, with specifics dependent on how the S-GW and P-GW are split. Relative to 2G/3G, it is expected to be more technically feasible and attractive to distribute EPC gateway functions. Conversely, there remains an argument for centralization, and the adage "don't distribute complexity" could have been written for just this situation. There is an opportunity to create a new "Access-Independent Gateway" product category that incorporates SAE Gateway functions. In line with the desire to further flatten the network and streamline operations, collapsing functions into a multiservice access gateway has significant potential over the medium and long term, despite near-term implementation and organizational challenges. Where wireless and wireline functions are required to be supported on the same physical equipment, edge-router platforms are likely to dominate.

1.2 Report Scope & Structure The report is structured as follows: Section II places the EPC in context, with an outlook for packet data services growth (subscriber adds, traffic volume, applications, operator revenues), a discussion of the primary functions of the packet-switched core, and a view on the ability of EPC to enable operator business objectives. Section III analyzes how the logical EPC network architecture, as defined by 3GPP, might be implemented in practice. It discusses integration with existing 2G/3G networks and the merits of overlay deployments versus a converged core. This section also looks at various deployment alternatives, such as the centralization or distribution of SAE Gateway functions, and how, over time, packet gateway functions may align (or overlap) with the underlying IP/MPLS network. Section IV provides an analysis of platform choices for EPC nodes, with an assessment of the functionality and performance of legacy hardware and its suitability for the next-gen packet core. It includes an analysis of EPC bearer-plane requirements and discusses the use of blade server platforms for control-plane and mobility management functions. Section V highlights the major issues that operators should consider when evaluating platform choices for EPC, and specifically for the S-GW and P-GW nodes.


II. EPC Market Requirements This section places the EPC in a wider market context. It examines the outlook for packet data services, discusses the primary functions of the packet-switched core, and offers a view on the importance of EPC to enabling operator business objectives.

2.1 Outlook for Packet Data Services The mobile phone is the fastest-growing consumer electronics product of all time, growing from virtually nothing in 1990 to more than 3 billion active connections and 400 million devices sold in 2008. Mobile data, while not yet in the same league, is now showing impressive growth. Active 3G connections (HSPA and EVDO) were in the region of 200 million at the end of 2008 and expected to grow to more than 1 billion by end of 2012. Mobile operator data revenues are growing at around 40 percent annually. Advanced 3G networks now provide typical downlink data rates of about 1 Mbit/s, and average busy-hour throughput per 3G (HSPA) subscriber is expected to increase from around 250 bit/s in 2008 to 115 kbit/s by the end of 2012. Viewed in this context, the long-term outlook for packet data services over LTE is positive. The chart below shows growth in mobile broadband connections by technology through 2012. We anticipate this strong growth in 3G (HSPA and EVDO) will prime the market for LTE, which will launch in 2010, add approximately 3.5 million connections in 2011, and then ramp up from 2012. Figure 2.1: Growth in Mobile Broadband Connections by Technology

In the longer term, LTE can be expected to gain greater scale than current 3G technology, due to the unprecedented international alignment behind LTE and through features such as spectrum flexibility, the unification of TDD and FDD operation, and the adoption of IP-centric network archi-tectures. Heavy Reading's aggressive forecast scenario sees more than 450 million active LTE connections in 2016, assuming a growth rate similar to 3G in the years following commercial launch and a starting point of 4 million subscribers in 2011. This growth in connections, average traffic per user, and demand for more sophisticated applica-tions should mean that the mobile packet core market, which was worth an estimated $1.4 billion in 2008, remains healthy and vibrant.


2.2 Functions of the Packet-Switched Core The packet-switched core is fundamental to wireless data networks. In LTE networks, the EPC manages mobility, tracks users as they move between cell sites, and routes traffic accordingly. It also serves to anchor mobility across access networks, as users roam between 2G, 3G, and LTE, or (potentially) seek to access "LTE services" from generic IP access networks such as home broadband or WiFi. It provides connectivity to desired network services and, because all of an LTE subscriber's data traffic must pass through the packet core on its way to the applications/service provider, EPC is critical to LTE services, with extremely high value placed on reliability and performance. The importance of the EPC is more significant in LTE than the packet core has been historically in 2G/3G networks, because it marks the first time within a mainstream 3rd Generation Partner-ship Project (3GPP) standards release that a mobile network can be considered an all-IP net-work. With no circuit-switched domain specified in LTE/SAE evolution, EPC is responsible for real-time services such as carrier-grade VOIP and video. This requirement for end-to-end quality-of-service (QoS), and by implication, to differentiate from best-effort services, is not unique to LTE, but represents the first time it has become a critical requirement in mobile networks.

2.3 Enabling Operator Business Objectives With EPC In addition to providing 3GPP-specified services, the EPC can also be seen as part of an opera-tor's "service engine" used to deliver differentiated services and applications. The intent is for an operator to extend its business outwards from basic connectivity into higher-value services that diversify the revenue and margin mix and – not to put too fine a point on it – expand the opera-tor's addressable market. The ability to differentiate services is already important to 2G/3G packet core networks and will become even more important in LTE. It is, however, an area that will need significant product in-novation on the part of the EPC equipment vendors if operators and users are to capture the full value of LTE access networks. While current core networks offer interfaces to the higher-layer policy control and do a reasonable job of enforcing this business logic, the outcome in terms of marketable services is often rather coarse-grained and lacks sophistication – for example, the technology is often used to block bandwidth-hogging applications, but less often used to capture extra revenue from these users. Part of the reason for this coarse-grained policy control is that end users are not yet receptive to services that, in essence, seek to charge higher prices for superior experience. It is still early days for mobile broadband, after all. Another reason is that, although technical mechanisms exist to implement sophisticated services, they are often not practical to implement in 3G and depend on hard-to-achieve alignment between technical and marketing divisions within an operator. However, this situation will evolve rapidly with the introduction of LTE: The ability to elegantly support differentiated services through core network technology will likely be a pillar of the LTE investment thesis at many operators. The fundamental point is that, as well as wanting to capture the lower cost per bit and superior data throughput of LTE, operators need to simultaneously generate increases in revenue per bit through advanced services alongside low-cost, best-effort Internet services – and that's where the challenge of EPC really lies.


III. Implementing the EPC Network Architecture This section examines the evolution of the 3GPP packet switch architecture from the 2G/3G core through to the EPC, highlighting major implementation decisions and challenges.

3.1 Logical EPC Architecture The EPC is defined in 3GPP Release 8 and is integral to the industry drive toward flatter, all-IP networks. There are two main components. The first, TS 23.401, is an enhancement to the exist-ing GPRS core network architecture intended to support the Evolved UTRAN (a.k.a. LTE RAN). It uses GTP as an underlying protocol. The second, TS 23.402, is an architecture enhancement to integrate non-3GPP access networks (such as CDMA) and uses Proxy Mobile IP instead of GTP. The high-level LTE/SAE architecture is shown in the diagram below. A major feature is the split between the control and bearer planes to allow each to scale independently. Specific EPC com-ponents are the Mobility Management Entity (MME) in the control plane and the Serving Gateway (S-GW) and Packet Data Network Gateway (P-GW) in the bearer plane. Figure 3.1: LTE/SAE Network Architecture & EPC

The Mobility Management Entity (MME) is a control-plane element that manages mobility in 3GPP access networks and carries out functions such as tracking users and making gateway selections. In effect, the MME controls terminal handovers. The Serving Gateway (S-GW) terminates the interface from the LTE radio access and is the local mobility anchor point for inter-eNodeB handovers and inter-3GPP mobility. Other functions include lawful intercept, and some charging and policy enforcement functions. The Packet Data Network Gateway (PDN-GW or P-GW) terminates the interface from the S-GW and connects to external packet networks. It provides the mobility anchor across non-3GPP access, interacts with the IMS service layer, and is a key node for policy enforcement. The Policy and Charging Rules Function (PCRF) is a control-plane element that is not, strictly speaking, an EPC element, but is required to give dynamic control over bandwidth, charging, and network usage.


These nodes are specified as logical network elements, but can be implemented according to operator or vendor preferences. The S-GW and P-GW could be combined into a single node, for example, or an MME function could be implemented on the same platform as an S-GW.

3.2 Integration of 2G/3G Packet-Switched Core With EPC The launch of GPRS in 2001 introduced packet switching to TDM-centric mobile networks and defined the network architecture that served as the foundation for 3G deployment, and which pre-vails to this day. The first GPRS packet core products were designed for low-rate data services and, initially, were based on sub-optimal platforms. When 3G was introduced a few years later, it was deployed as an overlay to meet the step change in product and network requirements. As 3G data service bedded down and dual-mode terminals became available, operators began to migrate 2G data users to a combined 2G/3G packet core. However, the transition wasn't without hitches. With the rise of mobile email (specifically BlackBerry), operators found that their highest-value data users were exclusively on 2G, prolonging the life of 2G equipment, while 3G traffic was sent over the state-of-the-art IP/MPLS networks to new-generation packet core nodes. Could the introduction of EPC follow a similar trajectory? Many of the same dynamics are in place: a step-change in data traffic and an equipment refresh are both anticipated. And, with data revenue now critical, there's a clear need not to risk 3G data service quality. This indicates that an overlay deployment for EPC will be the most frequent rollout scenario, with a migration to con-verged 2G/3G/4G packet core coming in a later phase. However, mobility and terminal interaction across LTE, 3G, and 2G networks will drive demand for tightly integrated core networks, even where EPC is deployed as an overlay. Unlike 3G, which was first introduced to a small base of mobile data users, LTE will be marketed to a large base of active data subscribers that will expect applications to work seamlessly between 3G and LTE. The effect will be to generate high volumes of inter-RAT (radio access technology) handovers, which will require tight integration between the EPC and existing 2G/3G network elements, such as RNCs, HLR/HSS, and SGSNs. Some of these interactions are shown in Figure 3.2. Figure 3.2: High-Level Interactions Between EPC & Legacy Packet Core


Due to this complexity, and to reduce cost, operators will ultimately want to operate common packet core equipment that supports both 2G/3G and EPC functions. To meet this need, some vendors aim to offer EPC via a software upgrade to equipment already deployed in the field. This could mean upgrading a GGSN to support P-GW functions via software, for example.

3.3 New Platform Requirements We are somewhat skeptical about the opportunity to upgrade deployed equipment via software to full EPC functionality, and instead believe that an overlay deployment will prevail in the majority of networks. This will minimize the risk of disruption to today's mainstream data services and allow operators to optimize LTE and EPC ahead of migrating to a converged 2G/3G/4G packet core. With existing packet core networks running at around 50 percent capacity, and with any head-room earmarked for the ongoing ramp in 3G data, the EPC overlay would logically utilize new equipment. This likely means new hardware platforms rather than reuse of 3G products. Within one to two years of commercializing and stabilizing the EPC, however, operators will want to consolidate wireless access networks. This is likely to mean migrating 2G and 3G to the new-generation packet core equipment, which will effectively turn EPC and the new-generation equipment into the foundation of the converged 2G/3G/LTE core network. It is at this point – when that mix of 2G, 3G, and LTE services has been proven to work – that software upgrades start to be appealing. If an operator's installed base of 2G/3G equipment was deployed recently, it could become attractive to reuse that asset as part of the converged core. The question then becomes one of operational cost: Does the legacy equipment have a cost of ownership that makes it worth retaining and upgrading? Or does the new EPC equipment gener-ate sufficient savings to encourage operators to migrate 2G/3G rapidly? Related to this are the potential opex advantages to selecting EPC equipment from the vendors that already supply packet core and/or IP networking equipment to the operator – for example, to retain common management tools and processes.

3.4 Implementation Options Practical implementation of the System Architecture Evolution (SAE) defined by the 3GPP presents numerous decision points for operators that must be realistic in the short term, yet retain the flexibility to align with a long-term vision for the network. In practice, this is likely to mean that EPC will be deployed in much the same way as 3G currently is, leveraging the same facilities and operational processes. In later phases, however, operators will need to take a view on their target network architecture and address how to scale EPC. This is where the uncertainty lies and where operators will take different decisions based on individual circumstances, refresh cycles, and product capabilities. Some of the key decision points are outlined below: Should the S-GW and P-GW be combined into a single node? Given the desire to simplify the architecture and flatten the network relative to 3G, a unified "SAE Gateway" is appealing. Yet be-cause S-GW and P-GW perform different functions and scale differently, it could be appropriate to deploy them separately (as is the case with SGSNs and GGSNs today). The P-GW provides the interface to external packet networks, and performs traffic analysis and policy functions, which are not required in the S-GW. The S-GW is linked more closely to the RAN, is focused on mobility functions, and interacts with large numbers of base stations. To what extent should "SAE Gateways" be distributed or centralized? This depends in part on the split between S-GW and P-GW, but relative to 2G/3G, it is expected to be more technically attractive to distribute packet gateways in the EPC to cope with anticipated LTE traffic, improve redundancy, and enable local Internet break out. Conversely, the adage "don't distribute complex-ity" could have been written for just this situation. Network size will be a key factor in the decision, with very large networks likely to have a mixed topology.


Where will MMEs be located? And which team will manage them? The MME is expected to be a centralized node – more so than an SGSN – but where it should be deployed is an open question. In the converged core, there may be a requirement to combine some MME and SGSN control-plane functions. This is challenging from some perspectives, however, and could impact deployment choices. How will the overall policy management architecture impact EPC? The EPC is responsible for policy enforcement, making interaction with policy control and business rules software critical. With more sophisticated QoS being used in LTE, it will also have to integrate more tightly with resource management mechanisms at the cell site, with the underlying IP transport layer, and with third-party application providers. How will transport network choices influence deployment? 3GPP defines an architecture that is largely independent of the transmission layer. In reality, however, the transport network architecture could have a significant impact on EPC – for example, where Layer 2 Carrier Ether-net dominates, a more centralized SAE Gateway may make more sense. Alternatively, where the network leans more toward Layer 3, there is also an opportunity to distribute gateways and per-haps integrate that capability with edge routers. Again the distinction between S-GW and P-GW may be important, with S-GW more likely than the P-GW to be integrated with the transport layer.

3.5 Alignment with IP/MPLS Core Networks This is an area where the choice of hardware platform for the S-GW and P-GW could potentially have an important impact. In the first instance, the EPC deployment model will be one of dedicated mobility nodes (S-GW and P-GW) installed on top of IP/MPLS core networks. In most networks, the IP core has ample capacity and functionality to support the introduction of LTE and absorb the ongoing growth of 3G traffic and support the discrete bearer-plane and control-plane model. In this scenario, the S-GWs and P-GWs could be equally based on either an Advanced Telecom Computing Architecture (ATCA) platform, an edge-router platform, or a specialist mobile gateway platform, assuming oth-er feature requirements are met. For most deployments, we expect this model to prevail over the medium term, not least because transport and packet core organizations are typically separate entities, with different planning horizons, priorities, preferences, and so on. This is especially the case where the operator is both a fixed and mobile service provider. Over the longer term, however, there is potential for a blurring of the boundaries between the IP network and the EPC "mobility layer," as functions such as packet classification, security, policy enforcement, etc., can be carried out in either domain. Moreover, in the scenario where gateways are deployed to the edge of the network (i.e., toward RNC or MTSO sites), the S-GW and P-GW applications could be supported on multiservice edge-router platforms that also operate as next-generation backhaul equipment in the metro network. Such a scenario faces challenges, but it does have some appeal in line with the goal of flattening the network to reduce costs, and several vendors are now positioning around the concept. Initially this deployment strategy could be adopted by smaller national or regional operators as they deploy LTE. For example, if the backhaul network needs to be refreshed and re-architected anyway, and the operator selects a multiservice edge-router product, it might make sense to add, say, an S-GW blade with a view to streamlining the overall network. Another area where this model could make sense is in turnkey buildouts where the vendor is contracted to a "build-operate-transfer" type deal with potential for ongoing managed services business. In this case, the onus is on the vendor to provide an end-to-end LTE network and to minimize the cost of op-erations and maintenance as much as possible. Clearly, such an implementation would not be suited to blade server platforms or other non-router platforms.


The counter view also holds weight: Rather than pushing Layer 3 routing and SAE Gateway nodes out toward the edge of the network ("don't distribute complexity"), the focus instead should be on low-cost Layer 2 Ethernet transport backhauling traffic to "super" EPC nodes, which are highly intelligent devices capable of advanced traffic management (traffic shaping, DPI, enforce-ment, etc.) from a centralized location. In this case, an ATCA blade server platform, for example, is potentially as suitable as an edge-router platform. In the broader sense, there is increasing alignment between the advanced service features sup-ported on IP networking equipment (e.g., edge routers) and the features required of S-GW and P-GW nodes. This may offer benefits in terms of being able to the leverage expertise and capability in the IP domain in the EPC – for example, IP subscriber management capability, enterprise VPN services, and hardware-based packet inspection.


IV. Platform Analysis for EPC Nodes This section provides an analysis of platform choices for EPC nodes, including an assessment of legacy hardware and its suitability for the next-gen packet core. It also provides an analysis of EPC bearer-plane requirements and how these influence SAE Gateway product design.

4.1 Economics of Platform Choices Decisions about which product platforms technology suppliers will use to support EPC applica-tions are informed by two key questions: What would be ideal for the application? And what other platforms do we have in the portfolio that would be suitable? Having tracked mobile packet core vendors' product evolution over several years, Heavy Reading believes leading suppliers have paid unprecedented attention to their platform choices for EPC. As well as creating an advantage that will enhance competitiveness across a vendor's LTE prod-uct portfolio, vendors also know they'll need to support EPC products for a decade or more. This requirement for long-term support tends to focus minds. The current mobile packet core market is relatively low-volume in terms of units – approximately 1,000 per year of GGSNs, SGSNs, and PDSNs – and, at around $1.4 billion per year, is not a huge market in absolute terms when split among vendors. Given the low product volume, sup-porting bespoke platforms is economically challenging for vendors. Even with ongoing growth in the mobile packet core market over the next five years – and Heavy Reading anticipates this segment will grow faster than wireless and telecom overall – this basic economic reality will pre-vail. EPC is a critical and strategic product area, but will remain a relatively low-volume market. Given this context, operators need to look hard at the underlying EPC platforms. Aside from the performance of the product itself, operators need to see how much support the vendor is putting behind the platform, what the shipment volumes are like, and what the longer-term roadmap and support outlook is. In essence, what is the vendor's depth of commitment to the platform? While the 2G/3G packet market shows there is some room for bespoke mobile packet gateway platforms, most vendors are likely to repurpose platforms for EPC that have already been used successfully in other telecom applications. In the bearer plane (S-GW and P-GW), for example, a tie-in with edge routers is a logical platform choice, allowing vendors to leverage scale and accu-mulated R&D resources. Operators should view this strategy as a positive, as it effectively delivers capability beyond what the EPC market alone could support financially. The danger is that EPC will not generate enough revenue to drive development of the underlying router platform or seriously influence its direction. The worst-case scenario is that the core business of selling routers takes precedence, leaving EPC innovation to wither. Nevertheless, we are yet to see a mobile packet core vendor with suc-cessful router products adopt anything other than a router platform for bearer-plane applications. Vendors that don't have a successful carrier-grade edge router in their portfolios obviously need to take a different approach to the EPC product platform. Historically the tendency has been to partner with router companies, and while this has been modestly successful, it is falling out of favor. Instead, vendors are looking to build products optimized for the mobile core using ATCA or bespoke non-router platforms. This drives a system design that leverages off-the shelf compo-nents, potentially including chassis, compute boards, processors, middleware, etc.

4.2 Vendor Strategies for EPC Platforms The size of the packet core market in the grander scheme has meant that products have, argua-bly, suffered from lack of investment and innovation. For big vendors, the revenues associated


with the packet core are not tremendously exciting, and while the gross margins are welcome, it's the strategic aspect of packet core – and the role it plays in pulling through business in the RAN, transport, and services domains – that makes it critical. This situation has meant that at each phase in the product refresh cycle, equipment vendors have tended to take the low-risk option and develop on legacy platforms. ATM switch and compactPCI hardware legacies, for example, live on in packet core products from several leading vendors. Vendors perceive the move to LTE and EPC as a major threshold, however, and are taking a long-term view of product requirements. There is broad acceptance that a new breed of IP-centric platform will be needed at some point in the product cycle, preferably for the start of LTE deploy-ments. Figure 4.1 summarizes the EPC platform strategies of seven leading mobile packet core equipment suppliers. Figure 4.1: Comparison of Platform Strategies for EPC

VENDOR EPC PLATFORM S-GW & P-GW PLATFORM

Alcatel-Lucent

• Introduce 7750-series router platform for S-GW and P-GW

• EPC applications on new "Mobility Services" blade • Introduce new ATCA platform for MME

Edge router

Cisco • Retain 7600 router platform for S-GW & P-GW • EPC applications on new "SAMI" blade

Edge router

Ericsson • Introduce Redback router platform for new S-GW and P-GW

• Retain proprietary platform used for SGSN, MME, and S-GW • Offer software upgrade to current PS core products

Edge router

Huawei • Introduce NE-series router platform for S-GW and P-GW

• Introduce new ATCA platform for MME • Offer software to some current PS core products

Edge router

Nortel • Introduce new ATCA platform for S-GW & P-GW

• Introduce new ATCA platform for MME ATCA

NSN • Introduce new ATCA platform for S-GW & P-GW

• Introduce new ATCA platform for MME ATCA

Starent • Upgrade current PS core products via software

• Uses bespoke hardware platform for all applications • Expected to introduce new platform within two years

In-house

Along with this broad-based transition to new platforms, a number of suppliers will also offer a software upgrade from 2G/3G to EPC for at least a proportion of their installed base. While we don't expect this to be the dominant way operators will introduce EPC (see Section 3.2 above), the offer of a software upgrade may mean operators can avoid having to write down asset values on legacy equipment prematurely. Some of the more modern equipment installed in networks would also be capable of supporting EPC for a limited time. GGSNs on the market today, for example, are capable of several Gbit/s of throughput per cabinet/rack – sufficient for the near-term load from LTE RANs, although perhaps not adequate over a five-year view. And as noted above, once software upgrades are in the mix, the question then becomes one of the operational cost of the existing equipment.


4.3 EPC Bearer-Plane Requirements While discussion of the suitability of different hardware platforms for EPC applications needs to be placed in the economic context discussed above (Section 4.1), there are nevertheless some requirements that any platform must take into account, including: throughput, latency, number of users, simultaneous bearers, and more nebulous on-board "intelligence." One way to position platform requirements for S-GWs and P-GWs is shown in Figure 4.2, which identifies the relative balance between control-plane and data-plane capability in packet gateway products. Mobile packet gateway products have historically traded control-plane capability against throughput, and are positioned in the top left quadrant. GGSNs and PDSNs, for example, have tended to focus on metrics such as number of simultaneous users (or active PDP/PPP contexts), rather than raw throughput. Wireline products, conversely, need to manage far fewer individual connections, with each connection generating far more traffic, and are placed at bottom right. Figure 4.2: Relative Balance of Control & Bearer Capability in Packet Gateway Products

This divergence in wireline and wireless products is unlikely to be fully bridged in the medium term, yet increasingly S-GW and P-GWs will need to evolve to meet new control- and bearer-plane requirements, as outlined here: Control Plane: S-GWs and P-GWs aggregate traffic from thousands of eNodeBs (base stations) and millions of subscribers, and therefore require greater control-plane capability than wireline products. The difference in EPC, relative to 2G/3G, is that there is no intermediate radio controller to aggregate traffic: Each eNodeB communicates directly with the S-GW. This architectural change alone increases control-plane requirements on the S-GW. When mobility is introduced to the equation, the signaling load (potentially) increases substantial-ly due to the need to set up, retain, and tear down bearers on the S-GW as users roam across the network. One illustration of the impact of mobility on control-plane capacity is already seen in


Direct Tunnel implementations, where the SGSN savings generated from bearer-plane bypass can be up to 100 percent greater in a low-mobility scenario than in a high-mobility scenario. The result is that SAE Gateway platforms, and specifically the S-GW, will need to continue to scale control-plane capability. This requirement is expected to be substantially more important than increases in raw throughput. Bearer Plane: Raw throughput is far from the primary challenge for mobile packet gateways, and while it may be a challenge for legacy products, it should not be a major concern for the new generation of S-GWs and P-GWs. Consider the difference between wireless data and fixed broadband access traffic volumes: Even accounting for the increase in LTE access speeds (relative to 3G/HSPA) and the smaller packet size associated with VOIP services, S-GW and P-GW capacity requirements will not get close to those offered by advanced edge-router products. More significant will be other bearer-plane ca-pabilities, such as latency, traffic analysis, and security.

4.4 Toward the Access-Independent Gateway Many of the advanced characteristics of wireless packet gateways are increasingly relevant to wireline products, or are now a common requirement across both domains. As vendors and oper-ators look to increase platform efficiencies and simplify networks, a requirement is emerging for a new product category that Heavy Reading has termed the Access-Independent Gateway (AIG). The AIG is essentially a common product platform defined in software for various applications, either operating in single mode (e.g., 2G/3G only) or multiple modes (e.g., an SAE Gateway and edge router). Key features are identified in the table below. Figure 4.3: Characteristics of an Access-Independent Gateway

BEARER PLANE CONTROL PLANE

Mandatory system design

requirements

Common resource management across bearer and control planes

High throughput (minimum 10 Gbit/s)

High transaction capacity (minimum 5,000 transactions per second)

System-level redundancy (fast failover and across blades with in-service updates)

Common hardware, software, and management resources across different

session and policy management features

Mandatory on-board features

Tunneling protocols

QoS (flow maintenance) Authentication and admission

Wire-rate packet classification

Mobility management (for wireless AIGs and converged AIGs)

Routing Standard interfaces to third-party session and policy management network elements

Aggregation

Additional on-board features

QoS (packet labeling) Traffic shaping

DPI (enforcement) DPI (policy)

Security (other)

Session border controller


The AIG is very much an emerging product category at this time, and our definition is not tied to any particular platform type; nor is it necessarily about using the same equipment for wireline and wireless access, although that is one example of an AIG. Another example would be a gateway that can function as a GGSN, SGSN, S-GW, P-GW, or ePDG; while exclusively a wireless prod-uct, this would still count as an AIG. In the EPC, the questions to be answered are: How specific are the requirements of S-GWs and P-GWs? Are they best met with upgraded edge-router platforms, or do they require bespoke plat-form development? If the latter, can the application justify that development commercially? The following section looks at some of the evolving bearer-plane requirements.

4.5 Evolving Packet Gateway Requirements – 3G to EPC With a majority of vendors saying they will introduce new or upgraded platforms to support EPC, it is worthwhile to consider the key differences between today's 2G/3G packet gateway require-ments and the EPC. Some of these are outlined in the table below: Figure 4.4: EPC Packet Gateway Requirements

REQUIREMENT IMPACT ON SAE GW

VOIP & Real-Time Services

• Ultra-low-latency packet processing

• Maintain throughput at 64-byte packet size • Ensure priority QoS handling

Latency • Sub-50µs "through-the-box" latency

• Minimal switching delay

Security • Terminate secure tunnels

• Requires dedicated crypto hardware

Throughput • Significant increase in traffic per subscriber

• Need 10s of Gbit/s and ability to scale further

Traffic Analysis

• Wire-rate packet classification

• DPI with minimal impact on system capacity • Heuristic analysis for encrypted services

• Upgrade libraries regularly (outside standard software release cycle)

QoS • Shaping, policing, and marking of traffic flows on a per-app/user basis

• Extend QoS classes into RAN and transport domains • Apply to QoS to third-party service providers

Perhaps the key difference is that the packet-switched domain must now support all services, including real-time VOIP, creating a need for ultra-low-latency packet processing. Even though packet core products have improved considerably over recent years, core networks still tend to induce undesirable amounts of delay to bearer setups, applications, and so on – whether due to the products themselves or the way the network is implemented. Given that latency reduction is a pillar of LTE and the flat network architecture, there is no tolerance for core network equipment to add delay to the service or session. A related element is the requirement to support advanced services, with wire-rate packet inspec-tion and traffic-analysis techniques (heuristics, etc.) as a starting point. Traffic analysis is critical to give operators an understanding of traffic flows and enable them to take action to assure that


business logic can be applied. This does not automatically imply blocking or degrading of applica-tions or users, although that is one application of the technology. How exactly packet inspection and DPI is implemented is important. Two points stand out: First, reduced performance in other parts of the system due to DPI being "turned on" is less acceptable than in the past, and a greater volume of traffic (and a greater proportion of traffic) will have to be analyzed in the EPC. Second, if DPI and traffic analysis is to be tightly integrated on, say, the P-GW node, there must be a way for vendors and operators to upgrade the libraries and rules engines more frequently than in the past. Another advanced requirement is the role the SAE Gateway plays in extending QoS management from a core network feature (as is the case today with 3G) to an end-to-end requirement. This is expected to work in both directions: Downstream, the EPC will have to enforce policy and allocate resources according to the prevailing radio access conditions; while upstream, it will have to inte-ract directly with large numbers of third-party providers, as well as the Internet, or corporate net-works. Developing end-to-end QoS is important to the LTE business case, and while technically possible in 3G, it is not yet practical to implement. These advanced bearer-plane requirements are the product of a more sophisticated, intelligent packet core, which in turn drives the need for greater control-plane capability on the S-GW and P-GW elements. For this reason, the ideal evolution from today's mobile packet core products is to increase packet-processing capability (on the x axis of Figure 4.2) while simultaneously enhanc-ing the control plane (on the y axis).


V. SAE Gateway Platforms & Operator Choices This section highlights the major issues operators should consider when evaluating platform choices for EPC, and specifically for the S-GW and P-GW nodes.

5.1 Edge Router, ATCA, or Specialist Platform? The platform on which S-GW and P-GW applications are based is not the deciding factor when operators select packet core suppliers, but one of many components to a complex decision. With pros and cons to all the products on the market, a certain platform type may be more appropriate to a particular vendor or operator, given their respective product and network refresh cycles. This "horses for courses" situation is likely to prevail with the move to EPC, not least because many of the early LTE deployments are expected to be end-to-end, with RAN and core provided by a single supplier in the first phase to ease the implementation. In this scenario, the RAN equipment is likely to dominate the purchasing decision, with the EPC secondary. This "RAN + core" purchasing pattern explains why the leading radio vendors roughly mirror their market share in packet core. Conversely, EPC can influence the competitiveness of a vendor's RAN portfolio – which, in a bit of circular logic, is why the underlying platform question is critical. The mobile core market today is split roughly 50/50 between vendors that offer packet gateway products based on edge-router platforms and those that use some kind of non-router platform, such as ATCA, compactPCI, ATM switch, or a bespoke platform. This bifurcation of platform strategies is likely to persist in the transition to EPC, albeit with router-based products gradually gaining share, due to the changing market-share picture in the wireless market as a whole.

5.2 Platform Heritage & Economics Perhaps more important than the straight choice between edge router or non-router platforms are the business motives that determine the vendor's choice of hardware platform for EPC and the extent to which they align with operator's own best interest. Some of the key questions to ask about SAE Gateway platforms are: How important is the SAE Gateway platform to the supplier's wider business? This is a fun-damental point. A platform that is orphaned within a vendor and only used in a single or few ap-plications is likely to suffer lack of R&D investment. This is especially the case at a time when vendors are consolidating platforms and streamlining development organizations. Conversely, if the platform is financially successful for the vendor and has a history of being refreshed, then it can be expected to secure a larger share of ongoing development resources. What are the shipment volumes of the underlying platform? Closely linked to the above issue of financial viability are the unit shipments of the underlying platform. As noted, the mobile packet core market as a whole is relatively low-volume, which strains the ability of vendors to support platforms, influence component suppliers, and so on. A platform that ships in volume for other, related applications is more likely to enjoy long-term support. Is the underlying platform an in-house technology initiative, or is it provided by a partner? Wireless vendors as a group have often partnered with specialist packet core vendors or used a partner's hardware platform to run their own application software. While there is still a place for this type of arrangement, especially where the operator has a best-of-breed purchasing strategy, using a third-party platform for an in-house product is decreasingly viable today. We expect Tier 1 vendors to migrate to fully-owned platforms which they control exclusively. What is the total cost of ownership (TCO) profile of the platform underlying EPC? There are multiple facets to this question, related to the age of the operator's installed base, the vendor's


product cycle, the availability of skills in the market, and so on. One area for operators to be wary of is that the product sale may be a vendor's stalking horse for services and integration work. This is of particular relevance to EPC vendor selection, because TCO will ultimately make the case for a converged 2G/3G/LTE packet core. How tightly are the EPC application software and hardware linked? This is a difficult point. Tight integration of application software and hardware enhances performance, but reduces flex-ibility. Loose coupling aides the portability of application software and increases the vendor's "platform agility," potentially increasing the frequency of product updates. With vendors adopting new hardware platforms for EPC, and intending to port 2G/3G applications to the new platform in the future, software portability is a valid question.

5.3 Edge-Router Platforms For CTO teams evaluating packet core evolution and for the executives and engineers with direct responsibility for this part of the network, there is a definite allure to edge-router platforms. Many of these people have backgrounds in IP networking and are drawn to the "router priesthood" – a trend that is reinforced in the move to EPC and all-IP wireless networks. At the same time, it is appreciated that mobile networks are different than wireline, with distinct product requirements. Moreover, it's obvious that not all routers are created equal, and that S-GW and P-GW applica-tions are not implemented equally well by each router vendor. Key questions to ask are: How exactly are EPC applications implemented on the router? In most cases, a dedicated module or blade is added to the router to provide EPC functions. The question then becomes, is this module/blade the S-GW or P-GW itself, with more blades added for extra capacity in a serial fashion? And is this a self-contained SAE Gateway, or would additional modules/blades be needed for advanced services, such as DPI or content charging? How, if at all, does the EPC product benefit from the router's inherent service capabilities? This basically comes down to which advanced services the router supports natively (e.g., packet classification, QoS control) and how these features can be used to enhance the S-GW or P-GW capability. Packet classification integrated in silicon or bearer QoS control would have relevance in a P-GW application. If this is already embedded as a feature of the underlying platform, it may provide performance advantages. How would the edge router support the addition of control-plane elements to the chassis? While it is anticipated that control-plane nodes such as MME and PCRF will tend to be deployed as distinct network elements (based in blade-server platforms), in some cases operators may want to add these functions to packet gateway products. For example, an MME may be appro-priately collocated on an S-GW platform in certain circumstances. Does the router platform align with the rest of your IP networking infrastructure, and will it allow tighter integration between EPC and the IP transport layer? Where the operator uses common equipment in the IP layer and for EPC, there could be substantial operational advantag-es (skills, maintenance, etc.). In the longer term, network architecture evolution could also trend toward a further flattening of the network and tighter integration with the IP layer, as discussed in Section 4.4.

5.4 ATCA or Specialist Platforms One widely held view is that mobile packet gateways are emphatically not routers. The require-ments in terms of mobility management, number of simultaneous users, and advanced services are so different that this justifies a different platform type. This has led some vendors to develop purpose-built platforms or press compactPCI or ATM switch platforms into service. More recently, a number of suppliers have adopted ATCA for their next-generation S-GW and P-GW products.


This is a clearly a viable platform strategy, and one that has proven successful in some cases. Some key questions to ask are: Do the packet-processing capabilities of these platforms meet the data rates and latency requirements of LTE and EPC? It is possible to build high-density, low-latency products on ATCA or proprietary "non-router" platforms, and there are examples of where this has been achieved for mobile packet core applications. Equally, there are examples in which performance and features have been marred by poor hardware choices. Without the IP networking experience inherent in routing vendors, operators considering non-router platforms must be comfortable with the depth of IP expertise and capability within the vendor. Will non-router platforms scale with changing LTE demands over a five- or ten-year span? Packet core product requirements evolved significantly in the transition from 2G to 3G, and a change of a similar magnitude can be expected for EPC. This forward-looking stance must be reflected in today's platform strategy. The risk is that product concepts originally designed to support low-rate 2G/3G services (e.g., content-based billing, MMS bypass, URL filtering, etc.) may not scale to support broadband IP functions. If ATCA is used for SAE Gateways, does the vendor have a broad commitment to this plat-form across product lines? If so, what is the evidence that it has been successful, or other-wise? Has the vendor chosen ATCA due to a lack of other options, rather than for its inherent properties? Operators should also question how well the vendor is able to manage issues such as updates to firmware or middleware by third-party suppliers, and what knock-on impact that could have in the product itself (e.g., in terms of testing cycles and software updates). This will partly be a function of how much of the product is based on commercial-off-the-shelf subsystems and how much is developed in-house. How will the selection of non-router platforms impact the evolution of the wider network architecture? A design assumption of ATCA-based or purpose-built SAE Gateway platforms is that mobile network applications (e.g., S-GW, P-GW, ePDG) will generally be deployed on a routed IP/MPLS network. This is a logical positioning in line with the classic "church and state" distinction between 3GPP functions and transport. Over the medium term, however, this reduces the opportunity to flatten the network and converge functions currently resident in distinct layers. This speaks to the AIG concept discussed in Section 4.4.


VI. Background to This Study 6.1 Original Research This Heavy Reading White Paper was commissioned by Alcatel-Lucent, but is based on indepen-dent research. The research and opinions expressed in the report are those of Heavy Reading and do not represent the official views of the Alcatel-Lucent.

6.2 About the Author Gabriel Brown – Senior Analyst, Wireless, Heavy Reading Brown's coverage at Heavy Reading focuses on wireless data networking technologies, including 3G/HSPA, WiMax, and LTE, and particularly on how these technologies impact the wider mobile Internet services market. Brown has covered the wireless data industry since 1998, previously as Chief Analyst of the monthly Unstrung Insider, published by Heavy Reading's parent company Light Reading. Before moving to Heavy Reading, Brown was additionally responsible for the overall editorial planning of Light Reading's entire line of Insider research newsletters. Prior to joining Unstrung, Brown was the editor of IP Wireline and Wireless Week at London's Euromoney Institutional Investor. He often presents research findings at industry events and is regularly consulted by wireless networking technology leaders. Brown is based in the U.K. and can be reached at [email protected].

6.3 About Heavy Reading Heavy Reading (www.heavyreading.com), a unit of Light Reading (www.lightreading.com), is an independent market research organization offering quantitative analysis of telecom technology to service providers, vendors, and investors. Its mandate is to provide the comprehensive competi-tive analysis needed today for the deployment of profitable networks based on next-generation hardware and software. Heavy Reading 32 Avenue of the Americas New York, NY 10013 United States of America Telephone: +1 212-925-0020 www.heavyreading.com www.lightreading.com

mailto:[email protected]�


http://www.lightreading.com/�


http://www.lightreading.com/�

hr epc platforms wp 0409

Documents