Flat IP Architectures For Mobile Networks - Unstrung Insider Report

VOL. 6, NO.
6, JUNE 2007
Flat IP Architectures for

Mobile Networks
TABLE OF CONTENTS
I.
Introduction: Flat Is Back
II.
Evolving Today's Networks

Drawbacks of Existing Network Architectures
Future Requirements
III.
Base Station Routers

UMTS Products
CDMA Products
IV.
UMTS Enhancements Direct Tunnel Architecture

Background to 3GPP's Direct Tunnel Approach
Direct Tunnel With Retained RNC
Direct Tunnel With RNC Control Plane
Direct Tunnel With Collapsed RNC
V.
WiMax Network Architecture

ASN Architecture
ASN Profiles
VI.
3GPP System Architecture Evolution

Enhanced UTRAN
Evolved Packet System
VII.
CDMA Enhancements Ultra Mobile Broadband
VIII. Company Profiles

Airvana
Alcatel-Lucent
Alvarion
Cisco Systems
Ericsson
Huawei Technologies
Juniper Networks
IX.
Conclusions
Motorola
NEC
Nokia Siemens Networks
Nortel Networks
Samsung Electronics
Starent Networks
Report Highlights
Advanced base stations
that integrate radio control,
header compression, and
encryption with IP/Ethernet
interfaces are key to the
emergence of flat mobile
network architectures
Base station routers are a
response to the need for
lower-latency 3GPP/2
networks, with AlcatelLucent, Airvana, and
femtocell players such as
Ubiquisys leading the way
Direct Tunnel Architecture
is emerging as the most
viable evolution for W-CDMA
macro networks; example
implementations include
Nokia-Siemens's I-HSPA
and Ericsson's SGSN
bypass initiatives
The WiMax Access
Services Network is the first
standardized IP-centric
mobile network architecture
establishing principles being
adopted across the industry,
but vendor interoperability
remains a challenge
Use of this PDF file is governed by
the terms and conditions stated in the
Subscriber License Agreement
included in this file. Any violation of
the terms of this Agreement, including
unauthorized distribution of this file to
third parties, is considered a breach of
copyright. Light Reading Inc. will
pursue such breaches to the full
extent of the law. Such acts are
punishable in court by fines of up to
$100,000 for each infringement.
I.
Introduction: Flat Is Back
With the wireless industry focused on the challenge of efficiently delivering mobile broadband
services, most of the attention is on which air interface will deliver the fastest connectivity to the
greatest possible number of users. But behind this debate on radio technology, there's an entire
back-end network infrastructure required to deliver applications. Here, at least, the choice is
simple: Adopt a flat, IP-centric network architecture. The idea is to decouple radio access from
core network evolution and implement a network architecture that offers substantially lower
latency and lower costs, while using less equipment and maintaining end-to-end quality of service
(QOS) and mobility.
The requirement for flatter mobile networks was set out by Verizon Wireless CTO Dick Lynch in a
TV interview with Wireless Week in November 2006: "As you look at a legacy network today,
they're very hierarchical; there are a tremendous number of protocols and protocol conversions,
there are multiple levels of switches that you send things through to get to the other end. In a
flatter IP network, you tend not to do that and importantly, most of the protocol conversions are
all gone We're looking at something where we actually deliver IP packets from the cell site, or
base stations, through a network to the base station at the other end."
What complicates the drive to flat IP mobile networks is the need to take into account a vast
installed base of highly reliable bought and paid for legacy equipment, as well as the applications
it supports. Additionally, most operators are internally organized to reflect the requirements of
classic network architectures, with distinct radio access and core networking personnel. For these
reasons, there are multiple approaches to migrating today's hierarchical networks to the flat, IPoriented mobile broadband networks of tomorrow.
This report examines flat architecture initiatives within the classic 3rd Generation Partnership
Project (3GPP) and 3GPP2 standards bodies, as well as those of newer organizations such as
the WiMAX Forum. It identifies five primary strategies for implementing flat architectures in mobile
networks:
Base station router development: products that overlay existing networks; devices today
primarily focus on micro-, pico-, and femtocell applications
Enhancements to existing 3GPP specifications: intended to support High Speed Packet

Access (HSPA) deployments and the move to the Evolved HSPA radio interface.
Worldwide Interoperability for Microwave Access (WiMax) access services network

(ASN): the emergence of the IP-based WiMax architecture has cranked up the pressure
on competing technologies to respond
Long-Term Evolution (LTE) and System Architecture Evolution (SAE): defining an all-IP
architecture evolution for the 3GPP community
Ultra Mobile Broadband (UMB) and Converged Access Network (CAN): defining an all-IP
architecture evolution for the 3GPP2 community
The report also analyzes the strategies and product roadmaps of leading equipment suppliers
looking to support operators in the disruptive move to next-generation mobile broadband
networks and take advantage of this opportunity to win market share.
Companies profiled in this report: Airvana Inc. (Nasdaq: AIRV proposed); Alcatel-Lucent (NYSE:
ALU); Alvarion Ltd. (Nasdaq: ALVR); Cisco Systems Inc. (Nasdaq: CSCO); Ericsson AB (Nasdaq:
ERIC); Huawei Technologies Co. Ltd.; Juniper Networks Inc. (Nasdaq: JNPR); Motorola Inc.
(NYSE: MOT); NEC Corp. (Nasdaq: NIPNY); Nokia Siemens Networks, a joint venture of Nokia
Corp. (NYSE: NOK) and Siemens AG (NYSE: SI; Frankfurt: SIE); Nortel Networks Corp. (NYSE:
NT); Samsung Electronics Co. Ltd. (Korea: SEC); and Starent Networks Corp. (Nasdaq: STAR).
UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING
II.
Evolving Today's Networks
Mobile operators' desire to migrate their networks to flatter architectures is driven by the need to
deliver lower-cost, higher-performance services.
A. Drawbacks of Existing Network Architectures

The migration should be seen in the context of today's network architectures designed for 2G and
3G voice and data services. Taking the 3G UMTS network as a starting point, the system
architecture was developed 10 years ago and is characterized by a hierarchical topology in which
end-user data is transported across different network elements in multiple formats, often requiring
protocol translation at each device. An overview of the 3GPP's Release 99 protocol stack is
shown below.
Figure 1: 3GPP Protocol Stack (Release 99)
Source: Unstrung Insider

Such an architecture obviously offers some benefits and has proven relatively well suited to past
and current requirements. Moreover, the choice has been dictated to some extent by the
demands of mobility, radio resource management (RRM), interoperability, and security.
However, with advances in technology, more experience in operating 3G networks, increasing
economic pressures, and customer demand for higher-performance applications, there is general
agreement in the wireless industry that today's hierarchical architectures are unsustainable.
To take the example of a data application over 3G, information would traverse at least four
network elements the base station, the radio network controller (RNC), the serving GPRS
support node (SGSN) and the gateway GPRS support node (GGSN) before reaching the
Internet or corporate network. There is also a high probability that the traffic would need to pass
through additional deep-packet inspection or content filtering servers (each adding latency)
before reaching the external network.
Such a scenario has several drawbacks:
The capex and, more importantly, opex required to support all these pieces of equipment
is relatively high and is not scaleable.
The architecture introduces delay into the application performance (which is inexcusable
in wireless networks, where the radio link absorbs most of the delay budget). Moreover,
voice over IP (VOIP) capacity in mobile wireless networks is linked to system latency.
The architecture is tightly linked to a particular radio interface technology and cannot
easily be adapted or scaled to meet the requirements of emerging "4G" wireless
technologies.
Although the above example uses terminology specific to UMTS networks, the issue is roughly
analogous in CDMA2000 networks, albeit less acute due to the absence of the SGSN in the
Evolution-Data Optimized (EV-DO) data path.
Figure 2: 3GPP2 Protocol Stack (EV-DO)
B. Future Requirements
It is notable that all new wireless systems being proposed, from WiMax to UMB and 3G LTE, are
based around flat network architectures that affect both the radio access network (RAN) and the
core network. In the RAN, the trend is to increase the capability of base station equipment, to
subsume much of the functionality of radio network controllers (RNCs) such as media access
control (MAC)-layer processing, RRM, airlink QOS, header compression, and encryption.
Over time, the RNC (and SGSN in 3GPP) disappears from the bearer path, although they may be
retained for control-plane function in some architectures. A new network element best described
as an access gateway takes on a role very roughly equivalent to a GGSN or packet data serving
node (PDSN) in the bearer path, acting as a mobility anchor point. It also performs functions such
as base station aggregation, bearer-path QOS, and policy enforcement.
Essentially, mobile networks are moving toward a basic two-box architecture base stations and
access gateways at the user plane. At a high level, most of the flat network architectures being
proposed are similar, even if specific implementations vary. Despite entrenched interests in each
standards group, there is remarkable agreement as to what the network should look like, shown
in Figure 3.
Figure 3: Generic Flat Mobile Network Architecture

This prospect is prompting some operators to consider ways to decouple radio interface evolution
(base stations) from the core network evolution, adopting an approach in which multiple types of
radio equipment can be added to the core network with almost plug-and-play ease.
"Unfortunately, core and radio technologies have always been tied together and, while I'm very
much satisfied with a core technology of one type, it may not today be hooked to a radio type I
want in the future," explained Verizon Wireless's Lynch. "So, one of the things we're starting to
look at now is, how can we take the best of a radio technology, and the best of a core technology,
and put them together, even if today they aren't coming out of the same standards body
"If we get the flat network, with IP coming out of the base station, and IP going through the
switching infrastructure, and IP at the other end, I can see an almost mixand-match capability
between core networks and radio technology."
On the equipment side, the concept of multiple access methods connected to a common core
network is driving the need for virtualization of application software, such that a vendor's product
should be able to support multiple applications on a common software and hardware platform.
This is anticipated to lower capex and especially opex.
So a CDMA operator introducing WiMax would look to support, say, the PDSN, Packet Data
Interworking Function (PDIF), and ASN gateway (ASN-GW) functions on the same equipment.
Similarly, a UMTS operator upgrading to LTE might look to support the serving gateway and
packet data network (PDN) gateway functions on the same platform as its GGSN or packet data
gateway (PDG).
III.
Base Station Routers
One way to flatten mobile network architectures is to combine network functions into integrated
pieces of equipment. This strategy has been used by several equipment companies for both
standard and non-standard wireless systems and has lead to the creation of a category of
products known as radio routers or base station routers.
The concept was popularized by a startup called Flarion that spun out of Bell Labs in 2000 with a
product that married an orthogonal frequency-division multiplexing (OFDM) air interface with
integrated RRM and mobility control. This "radio router" product offered an IP/Ethernet interface
to the network and did not require radio controller or mobility gateway nodes.
While ultimately unsuccessful as a standalone company due to a lack of standards and a
supporting ecosystem (Flarion has since been acquired by Qualcomm), the base station router
concept has been picked up by several other vendors. The twist, typically, is that it has been
applied to standardized UMTS environments and targeted at specific deployment scenarios such
as in-building coverage using femto- and picocells, or capacity in-fill using microcells. Products on
the market today have not generally targeted mainstream macrocell applications, although this is
envisaged for the future.
A. UMTS Products
One supplier of UMTS base station equipment that integrates RNC, SGSN, and GGSN
functionality is Alcatel-Lucent, which perhaps unsurprisingly, given its Bell Labs heritage is the
lead proponent of this technology with its BSR product line. Figure 4 illustrates the differences
between a traditional hierarchical UMTS network and a flatter base station router network.
Figure 4: Today's UMTS Network vs. UMTS With BSR
Source: Alcatel-Lucent
In the traditional architecture depicted on the left, adding a new Node B necessitates RNC and
SGSN configuration changes and, potentially, capacity upgrades. In contrast, adding a new BSR
to the network on the right has limited effects on other nodes due to the removal of network
hierarchy and offers a potential 30 percent reduction in capex, says Alcatel-Lucent. The
concept of "collapsed RAN" products has also been adopted in the UMTS market by femtocell
suppliers such as Ubiquisys and in the enterprise market by Motorola, with its AXPT 3G picocell.
It is important to note that, although these base station routers integrate functionality from the
access and core network, they are still standards compliant and continue to support the RNC,
SGSN, and GGSN functions required from a handset perspective.
The challenge with this product category really comes from two areas: RRM (which affects handoff, throughput, uplink performance, and QOS) and core network integration. Meeting these
challenges requires bespoke engineering in the case of core network integration and proprietary
technology in the case of RRM.
The issue of base stations with autonomous radio management is an interesting one, with the
challenges of achieving high-level performance often being swept under the rug. In radio
networks that feature macrodiversity (in which the handset communicates with more than one
sector at a time) such as CDMA and Wideband CDMA (W-CDMA) distributed radio
management is more challenging and possibly not worthwhile, as RNCs already do a good job.
Sacrificing macrodiversity would reduce uplink performance and require operators to deploy more
base stations, incurring substantial expense. In orthogonal frequency-division multiple access
(OFDMA) radio networks, where macro-diversity is typically not used, such as WiMax and
upcoming LTE networks, distributed radio management is more viable.
On the core network side, many operators are uncomfortable with products that integrate GGSN
functionality because they complicate the application of network-wide policies implemented in the
GGSN, such as content filtering, billing, and traffic shaping. Resistance to GGSN integration is
also attributable to the fact that a fully integrated product cuts across the divisional structures of
many operators. As a result, suppliers are moving to offer products with a Gn interface between
the base station router and GGSN. Suppliers that can overcome these radio and core network
challenges will likely have an advantage in the move to next-generation wireless networks.
One interesting twist on the fully integrated base station router concept is the "Direct Tunnel"
architecture standardized in 3GPP Release 7 and discussed in Section IV.
B. CDMA Products
Without an SGSN node, CDMA EV-DO networks are already closer to a flat IP design than UMTS
networks. Nevertheless, there appears to be an opportunity for base station routers in this
market.
Airvana recently became the first prominent vendor to announce a set of base station router
products for the CDMA market, with a design that integrates the RNC and PDSN function into the
base station. The company expects to offer this through both its direct sales channels and original
equipment manufacture (OEM) channels. Given that Nortel and Alcatel-Lucent are already OEM
partners, it's possible that one of them could offer products based on the Airvana technology,
although this has not been confirmed by any of the parties involved. Alcatel-Lucent has said it
plans to introduce CDMA base station router products in 2008 without giving further detail.
There are also some CDMA femtocell players pitching this collapsed architecture for CDMA, with
Samsung and AirWalk Communications to the fore. Interestingly, AirWalk integrates the RNC
without affecting the PDSN, thereby avoiding the core integration issue discussed above.
The bottom line for base station routers is that, although they've not yet had a significant impact
on the market, the product concepts are informing decisions about how mobile network
architectures will be designed in future. Many features of these products, particularly RNC
integration, are being formally standardized through bodies such as the IEEE 802.16 Working
Group on Broadband Wireless Access Standards, 3GPP, and 3GPP2.
For more on femtocells and the move toward flat RAN network architectures, see the Unstrung
Insider report 3G Home Base Stations: Femto Cells & FMC for the Masses (Vol. 6, No. 1,
January 2007).
IV.
UMTS Enhancements Direct Tunnel Architecture
A. Background to 3GPP's Direct Tunnel Approach

Perhaps prompted by the proliferation of vendor-specific attempts to flatten the classic 3G
network, the 3GPP has moved to standardize a flatter RAN. The architecture and its various
options are discussed in the March 2007 Technical Report entitled Scope of Future FDD HSPA
Evolution (TR 25.999). Referred to as the Direct Tunnel architecture, it is part of the 3GPP
Release 7 specifications. As the name implies, the intention is to specify a low-latency
architecture suitable for HSPA radio technology, and particularly the HSPA Evolution functionality
specified in Release 7 that will increase over-the-air data rates through the use of higher-order
modulation and multiple-antenna schemes.
Although not officially noted by 3GPP, the Direct Tunnel architecture appears to be partly
influenced by the Internet HSPA (I-HSPA) architecture proposed by Nokia in 2005. This proposal
suggested developing 3G base stations with integrated RNC functionality, including integrated
RRM, header compression, and encryption. The base stations would then be connected over a
Generic Routing Encapsulation (GRE) or GPRS Tunneling Protocol (GTP) tunnel to the GGSN
node via the Gn interface. The architecture would eliminate two nodes in the network the RNC
and SGSN to deliver a lower-latency, lower-cost network. I-HSPA was conceived as suitable
either for greenfield mobile broadband deployments or as an overlay to existing 3G networks,
primarily for data services.
The 3GPP's Direct Tunnel architecture employs similar concepts, but with some important
differences. Notably, the SGSN is retained as a control-plane node (it acts as an authenticator),
and a modified RNC is retained in some of the options discussed in TR 25.999. The use of GTP
tunnels is also retained, since this feature is carried forward in the 3GPP Release 8 specifications
for SAE, discussed later in this report. There are likely multiple reasons for these changes. At a
competitive level, there would inevitably have been 3GPP members who would not want to see a
Nokia-Siemens initiative green-lighted through the standards process. And, at a practical level, it
is clear some operators want more flexibility in how the flat architecture is implemented.
Several operators interviewed for this report expressed strong interest in the Direct Tunnel
architecture, but wanted to implement it with as little impact on the RAN as possible (i.e. they
wanted to retain the RNC) because of the amount of equipment already deployed in the field and,
partly, because of the internal organization structure of the operator, with larger providers' core
and access networks often managed separately.
A feature known as carrier sharing means that Evolved HSPA and regular HSPA can be
supported on the same transceiver, with traffic split to the appropriate user-plane path. There is
also a requirement for inter- and intra-system handover. A detailed description of the architectural
options is presented in TR 25.999, which discusses three alternatives to the current architecture:
Direct Tunnel with retained RNC
Direct Tunnel with RNC control plane
Direct Tunnel with collapsed RNC
B. Direct Tunnel With Retained RNC

In the architecture shown in Figure 5, the functional split between RAN and core network is
retained, enabling reuse of the proven, interoperable Iu interface. Only the functions that reduce
latency and increase bit rate have been moved from the RNC down to the Node B. In this specific
variation of the Direct Tunnel architecture, user-plane traffic takes the light green Iu-packetswitched path directly to the GGSN, while the control plane takes the dashed red path via the
SGSN. In an alternative option, not shown, the SGSN can also be in the bearer path, although
this would seem to defeat the purpose of the proposed architecture.
Figure 5: Variation on Direct Tunnel Approach in 3GPP RNC Retained
Source: 3GPP TR 25.999 (March 2007)

This option appears least disruptive to existing networks and aligns well with current topologies,
with ciphering, RRM, and header compression retained in the RNC. However, it would seem to
lack some of the cost benefits of eliminating the RNC entirely.
C. Direct Tunnel With RNC Control Plane

The diagram below shows an architecture designed to address the limitations of the RNC as a
bottleneck for user-plane traffic. The dark and light green lines show two options for sending userplane traffic directly from the Node B to the core to reduce latency. Control-plane traffic follows
the path shown by the dotted red line.
Figure 6: Variation on Direct Tunnel Approach in 3GPP RNC Control Plane

In this variation, the RNC is retained only for control-plane functions, which helps in coordinating
mobility and inter-cell handovers. This allows the RNC control plane to scale independently of the
user plane. It also makes the Node B simpler than it would be if all RNC functionality were
integrated, but it creates challenges with respect to macrodiversity.
D. Direct Tunnel With Collapsed RNC

In this architecture, the RNC functions are integrated in the Node B (shown here as an Evolved
HSPA Node B), which has a direct user-plane connection to the GGSN in the core network, with
the SGSN retained for control-plane traffic. Communication between Node Bs takes place over
the Iur interface normally used for inter-RNC traffic.
One challenge for this architecture is inter- and intra-system call handover, which is supported
through a feature called "circuit switch-enabled handover." This architecture probably delivers the
greatest benefit over the existing UMTS terrestrial RAN (UTRAN), but it is also the most
disruptive, and therefore best suited to greenfield HSPA deployments or HSPA overlays for
packet-switched applications.
Figure 7: Variation on Direct Tunnel Approach in 3GPP Collapsed RNC

Supporting macrodiversity in this architecture would likely be difficult and would require a highspeed transmission network itself a challenge. It can be assumed, however, that Evolved HSPA
cell sites would be connected via high-speed links.
Also interesting is that this high-level architecture maps more closely to the proposed SAE
architecture being developed as part of 3GPP Release 8 than do the two previous options. This
has implications for vendor product strategies, possibly permitting the consolidation of R&D
investments.
10
V.
WiMax Network Architecture
The WiMax network architecture is far simpler than its 3G counterparts, and the mere existence
of WiMax companies claming to be developing all-IP mobile wireless networks has put a rocket
under the 3G community. In this sense, it can be argued that WiMax represents the driving force
behind the move to flat mobile network architectures.
The simplicity of the WiMax architecture relative to 3G is partly a function of WiMax being a
newer technology, developed with all the lessons of 3G in mind and none of the backward
compatibility challenges or legacy baggage. Inevitably, however, the complexity of WiMax will
increase as it is deployed on a large scale and as operators in different global markets demand
more features.
A. ASN Architecture
The architecture itself comprises just two nodes in the bearer path: the base station and the ASNGW. Together, these make up the ASN. The other notable elements are the mobile devices,
obviously, and the core services network (CSN).
Each logical element in the network is connected via a standardized "reference point" that defines
protocols and procedures to allow equipment to interoperate. Physical interfaces and transport
technologies are not standardized. The reference architecture was developed in the WiMAX
Forum's Network Working Group (NWG) and is described in the Mobile System Profile Release
1.0, which was published in February 2007. An overview of the architecture is shown in Figure 8.
Figure 8: WiMax Network Architecture
Source: Intel Corp.

Like the 3GPP and 3GPP2 standards, the WiMAX Forum NWG does not specify physical pieces
of equipment but, rather, logical functions that can be incorporated as vendors see fit. The base
station, of course, is normally a physical device that implements physical-layer and MAC
functions, and in some cases may also implement ASN-GW functions.
11
Often, however, the ASN-GW is a standalone device that supports both bearer-plane and controlplane functions via its enforcement point and decision point, respectively. Control functions
include mobility management, charging and authentication, and possibly RRM. Bearer functions
include user-plane forwarding and policy enforcement, and likely vendor-specific functions such
as packet inspection and traffic shaping.
Figure 9: ASN-GW Elements & Interfaces
Source: Nokia Siemens Networks

The ASN-GW typically connects to multiple WiMax base stations on the radio side and interfaces
with network elements, such as authentication, authorization, and accounting (AAA) servers,
home agents, and Dynamic Host Configuration Protocol servers, on the network side via the CSN
interface. The CSN is not an item defined in the architecture as such, but a collection of functions
found in any telecom network.
Given such a flat network, the way in which WiMax handles RRM and mobility is interesting, since
these are partly responsible for 3G's complexity. In a nutshell, it appears that WiMax sacrifices
performance for simplicity in these respects. This makes it likely that, over time, vendors will
introduce proprietary features to enhance performance.
The architecture offers three tiers of mobility management:
Intra-ASN: between base stations connected to the same ASN-GW
Inter-ASN: between base stations connected to different ASN-GWs
Inter-CSN: between ASNs connected to different core networks
B. ASN Profiles
The WiMAX Forum NWG's specifications offer three ASN profiles, partly defined by the way in
which RRM is handled:
Profile A: RRM split between the base station and the ASN-GW
Profile B: ASN integrated in the base station
Profile C: RRM integrated in the base station, with a separate ASN-GW
Profile C, a distributed architecture with distinct base station and ASN-GW nodes, is expected to
be the most popular of the three options. In this profile, because RRM is integrated with the base
station, the market for ASN-GW products is open to non-radio vendors, particularly suppliers of IP
networking equipment. This openness is expected to increase innovation and competition among
suppliers and is in contrast to current 3G architectures where, due to the tight integration between
base stations and radio controllers, operators must buy RAN equipment from a single vendor.
12
How open this ecosystem will prove in practice, however, remains to be seen. The reference
points specified in Release 1 are not subject to formal interoperability testing. And, as with 3G,
private agreements between vendors will be required to drive true interoperability. The WiMAX
Forum does not currently plan to formalize this interoperability testing (as it has done for the radio
interface profiles), but may do so in the future.
Although the WiMAX Forum takes the view that fine-grained RRM is not worth the added
complexity, several large vendors, citing performance benefits, support Profile A the
architecture that most harks back to the traditional RNC concept used in 3G. Therefore, the NWG
is considering options for more granular RRM in Release 1.5 of the specifications. This introduces
the risk of "feature creep" at a time when the 3G communities are simplifying their networks.
Figure 10: Summary of ASN Profiles
Profile
Description
Pros
Cons
Profile A
Centralized platform;
RRM split between base
station and ASN-GW
Supports soft handover
Interoperability difficulties
between base station and
ASN-GW vendors
Profile B
Distributed platform;
integrated base station
and ASN-GW
Simpler, lower cost for

smaller deployments
Complex, more
expensive for larger
deployments
Profile C
Distributed platform;
RRM in the base station;
separate ASN-GW
Easier to use different

vendors' base stations
and ASN-GWs
Soft handover is difficult;

requires signaling
between base stations

ASN-GWs will be available in multiple sizes, from a software implementation on the base station's
network card, to 1U cell-site boxes, to equipment based on blade server or edge router products.
WiMax suppliers appear to be working under the assumption that ASN-GW capacity of around 60
Mbit/s per base station will be required, given a 1/1/1 configuration with 10MHz channels (i.e. 20
Mbit/s per 10MHz carrier). It is also intended that ASN-GWs of different sizes will not be deployed
in a hierarchy, but will instead form a pool of gateways that retains the flat architecture. There are,
however, some moves to design a hierarchy of ASN-GWs to support services such as Multicast
Broadcast Service, a feature being defined in the WiMAX Forum NWG's Release 1.5 of the spec.
13
VI.
3GPP System Architecture Evolution
To remain competitive over the longer term, the 3G community is developing a new set of radio
access and core network specifications through the 3GPP. These will appear initially in Release
8, scheduled for completion in 2008.
A. Enhanced UTRAN
On the radio access side the initiative, commonly known as Long-Term Evolution (LTE), is aiming
for improved spectral efficiency, greater user throughput, and lower latency. It is also targeting a
simpler RAN that will dispense with RNCs by pushing functionality, such as radio control, header
compression, and encryption, into the base station (called an Enhanced Node B or eNB).
Note that while LTE has become a common industry acronym, in May 2007 the 3GPP adopted
the term "Enhanced UTRAN" in place of LTE. Figure 11 shows the high-level RAN architecture.
Figure 11: E-UTRAN Architecture
Source: Source: 3GPP (TS 36.300, Draft, April 2007)

The key elements are the eNB and the mobility management entity (MME)/SAE gateway. Unlike
previous 3GPP architectures, the base stations (eNBs) can connect to one another directly via
the X2 interface. The connection to the core is via the S1 interface, which offers the potential for
eNBs to connect to multiple MME/SAE gateways, enhancing mobility performance by reducing
the number of inter-gateway handovers.
B. Evolved Packet System

On the core network side, through an initiative known as SAE, or the Evolved Packet System
(EPS), the target is a low-latency, all-IP core able to support real-time packet-switched services.
Notably, EPS does not call for separate voice and data core networks, as are in use today and
being designed to support VOIP services. The new core network will, however, interoperate with
existing 3GPP networks and with non-3GPP access technologies.
14
The Enhanced UTRAN (E-UTRAN) shown above connects directly to the EPS core network. EPS
consists of two primary nodes: the Mobility Management Entity (MME), a control plane node; and
the EPS/SAE Gateway, which is primarily a user-plane node. It is expected that the physical
manifestation of the MME node will be as a blade deployed alongside the EPS Gateway, or it
may be implemented on 3G SGSNs.
The overall architecture is shown in Figures 12 and 13. In Figure 12, the serving and PDN
gateways are integrated.
Figure 12: Non-Roaming Architecture for 3GPP Access (Single Gateway Configuration)
Source: 3GPP (TS 23.401, Draft, May 2007)

Although this would appear to run contrary to the desire for flat networks, there is also an option
to separate the serving and PDN gateways, shown in Figure 13.
Figure 13: Non-Roaming Architecture for 3GPP Access (Separate Gateways Configuration)
Source: 3GPP (TS 23.401, Draft, May 2007)

An explicit requirement of the EPS architecture is to connect non-3GPP access networks, such
as WiMax and wireless local-area networks (WLANs). It does this, essentially, by building on the
Intelligent WLAN (I-WLAN) specification (TS 22.234). Broadly speaking, the I-WLAN security and
authentication model will be retained in SAE, with the primary difference being the way in which
mobility across networks is handled.
Note also the introduction of the concept of trusted access to the core via non-3GPP access
methods. One suggestion is that this could be used in WiMax networks that have acceptable
security mechanisms in place. The non-roaming architecture is shown in Figure 14.
For more on I-WLAN and the role of the PDG shown in Figure 14 as an enhanced PDG, or
ePDG see the Unstrung Insider report Mobile Network Security: The Threat of Convergence
& IMS (Vol. 5, No. 5, May 2006).
15
Figure 14: Non-roaming Architecture for Non-3GPP Access
Source: 3GPP (TS 23.402, Draft, April 2007)

The timeframe in which the E-UTRAN and EPS initiatives will have a commercial impact is hard
to judge. Given the weight of support behind 3GPP from the world's major operators and vendors,
it seems likely that this will ultimately emerge as the dominant architecture for mobile networks.
Current thinking is that the LTE/SAE specifications will be largely complete in 2008 and ready for
technology trials in 2009, with a view to possible deployment from 2010 or 2011. However, such a
schedule seems optimistic, considering that the SAE work in particular is behind schedule.
This puts the 3GPP process substantially behind the timelines of the WiMAX Forum and the
3GPP2 (discussed in Section VII). This may not be such a problem, since the 3GPP process is
nothing if not thorough. Note also that, in practical terms, carriers are preoccupied by developing
and monetizing their HSPA networks, allowing plenty of time for LTE/SAE to mature.
At one point, it was proposed that the SAE core network could be used to support UTRANs
based on the HSPA Evolution air interface. This idea was swiftly dropped due to handset
dependencies and the difficulty of supporting legacy devices. Current thinking holds the Direct
Tunnel approach to be the best and most timely way to deliver a flatter network architecture for
HSPA Evolution.
16
VII. CDMA Enhancements Ultra Mobile Broadband

The CDMA community is also pushing for a flatter network architecture through the 3GPP2. The
primary initiative is the Converged Access Network (CAN) architecture that is part of the UMB
specification process, previously known as EV-DO Revision C. Unlike 3GPP, the 3GPP2 does
not have plans for interim specifications to flatten the existing architecture. There are two reasons
for this: First, 3GPP2 networks already have fewer nodes than their 3GPP counterparts (no
SGSN, for example); second, the timeline for the "Beyond 3G" specifications is much shorter.
The first release of the UMB radio interface was published in April 2007, and specifications for the
evolved RAN architecture and the overall CAN architecture are scheduled for completion in
October 2007. This is a far more aggressive and compressed schedule than 3GPP's and upholds
the CDMA community's reputation for moving through the standards process quickly. With UMB
chipset samples scheduled for the end of 2008, limited commercial services could begin in 2010.
The overall system architecture is shown in Figure 15. At a high level it is similar to the WiMax
ASN and the 3GPP's E-UTRAN. The two primary nodes are the Evolved Base Station (shown
here as eBR) and the Access Gateway (AGW).
Figure 15: EV-DO Architecture Evolution With UMB
Source: Alcatel-Lucent
The eBR handles all radio-related functions, as well as features such as header compression,
encryption, and over-the-air QOS. The AGW acts as the mobility anchor and takes on both
bearer- and control-plane functions, such as charging and authentication, user-plane forwarding,
and policy enforcement, likely in addition to vendor-specific functions such as packet inspection.
The primary challenge for UMB is building a strong ecosystem around it, given that the CDMA
market as a whole appears to be a little on the back foot. While leading CDMA vendors, AlcatelLucent, ZTE, Nortel, Motorola, and Qualcomm have committed to UMB publicly, there are
concerns about the real level of R&D commitment to this technology by some of these companies
in view of the fact that no major operator has yet committed to deploying UMB, or publicly
expressed a strong desire to do so. Operator involvement to date has been limited to
specification development.
17
VIII. Company Profiles

A. Airvana Inc.
Based in Massachusetts, Airvana (Nasdaq: AIRV proposed) is a privately-held company
specializing in CDMA radio access systems. Since its inception in 2000, the company has been a
firm believer in all-IP mobile networks and has long promoted the IP RAN concept.
Its primary sales channel are OEM partners, most notably, Nortel Networks, to which it supplies
EV-DO base station channel cards. It has around 490 employees, of which 100 are in India. In
April 2007, the company announced its intention to IPO. The S1 statement shows fiscal 2006
revenue of $170 million, with product and service billings of $141 million, of which Nortel
accounted for 95 percent. Cash flow from operating activities was $25 million, with operating
income of $57 million. It spent $55 million on R&D over the year.
The company also noted in its S1 that in October 2006 it signed an OEM agreement with AlcatelLucent to license its EV-DO software. It did not specify for which product types.
Base Station Router: In March 2007 Airvana announced its plans for base station router
products for the CDMA EV-DO market. The intention is to integrate RNC, and PDSN functionality
into the base station to enable operators to deploy flat RANs. The company says the
announcement is reflective of more than four years work on the concept and that the first OEM
products based on the flat architecture RAN are expected to begin market trials in early 2008.
Airvana's flat RAN architecture is based on software modules that can be ported to proprietary or
standards-based hardware platforms. Initial implementations are likely to target enterprise,
campus, hot spot, and public safety, applications where macro network coverage is challenging
and where Internet backhaul is required. Later implementations will target macro network
applications and the company says there's no technical reason to prevent this being successful.
The collapsed RAN concept is also being applied to the company's femtocell project. To
accelerate this project, Airvana acquired U.K. startup 3Way Networks for $11 million plus stock in
May 2007.
Direct Tunnel Architecture: Airvana is not active in this market, although is likely to offer PDG
(3GPP) and PDIF (3GPP2) applications based on its Universal Access Gateway product for
femtocell, WLAN, and other IP network integration.
Airvana hasn't commented on its activities relative to SAE/LTE, due to its IPO quiet period. The
company has said it doesn't anticipate entering the WiMax market, either on the radio access
side or with an ASN-GW.
UMB: As a CDMA stalwart, it would seem natural for Airvana to continue down the 3GPP2
evolution track. However, as with other vendors in this market, it appears to be taking a wait-andsee approach. Airvana's core strength is in CDMA radio rather than OFDMA. It has commented in
its S1, however, that it anticipates participating in the "4G" wireless market in some form.
Unstrung Insider Comment: Having promoted the flat all-IP network architecture since its
inception as a company it must be gratifying for Airvana to see the market finally moving in its
direction. The challenge, obviously, is to turn this into business.
The company's strategy of moving into the larger 3GPP and convergence markets via its
femtocell strategy makes sense, but it's a tough road to travel. For femtocells, in particular,
Airvana needs to show it can deliver an end-to-end solution very quickly if it's to succeed in
selling direct into major 3G operators. On the base station router side, it has an opportunity to
establish leadership in the CDMA market.
18
B. Alcatel-Lucent
The merged Alcatel-Lucent (NYSE: ALU) has the largest portfolio of mobile wireless infrastructure
equipment of any supplier in the world and is unique in having a strong position in UMTS, CDMA,
and WiMax. This gives it an unequalled opportunity to influence the direction of the global
wireless market. In theory, at least, the move to IP networks should provide a huge incentive for
the company to consolidate its product platforms. However, streamlining multiple product lines
without compromising customer-facing activities is a daunting task.
Base Station Routers: As an independent company, Lucent had championed the base station
router concept since at least 1996, with its Bell Labs research division acknowledged as a leading
authority with particular expertise in distributed RRM (a capability that underpins flat mobile
networks). Lucent launched its BSR product line for UMTS networks in 2006, with devices that
integrate RNC, SGSN, and GGSN functionality into the Node B. The BSR was initially offered for
pico- and microcell applications, and Lucent claims to have had success with this concept in
several operator trials across Europe.
Alcatel-Lucent remains committed to the base station router concept and views "maximally flat"
architectures, coupled with the IP Multimedia Subsystem (IMS) control layer, as the most
desirable long-term network architecture. It describes the BSR as a software architecture, which
could be ported to the UMTS products acquired from Nortel. It says certain UMTS operators are
asking for macro BSRs in the 2008, a timeframe it anticipates being able to support. The
company says it also planning 2008 availability of CDMA versions of the BSR, which will be
available in multiple form factors rather than just femto-or picocells.
Direct Tunnel Architecture: Despites its support for maximally flat networks, Alcatel-Lucent is
aware that there are practical challenges to this architecture for carriers with an installed base of
3G equipment. For this reason, it is also developing base stations with collapsed RNC
functionality, with either an Iu interface to the SGSN or a Gn interface to the GGSN. The
challenge here will likely be aligning the base station router concept and experience with the
UMTS products acquired from Nortel. It's unclear what exactly the schedule or strategy is for this.
The first product designed and built since the merger of Alcatel and Lucent, and incorporating
technology acquired from Nortel, is expected to be the fully flat Femto BSR, offering multiple core
network integration options, including Iu and Gn interfaces, and planned for commercial launch in
the first half of 2008.
E-UTRAN & EPSs: On the base station side, the groundwork for LTE has been laid with the BSR
products, and specifically on the OFDMA air interface, with the WiMax products. On the packet
core side, it would seem logical that the company will base future development on its internally
developed Intelligent GGSN, or iGGSN platform. The Alcatel-Lucent SGSN is the first Advanced
Telecom Computing Architecture (ATCA)-based SGSN on the market to ship in volume and the
iGGSN seen as a key strategic step in the evolution to an access-independent IMS network,
serving as the policy enforcement point across multiple access networks.
WiMax ASN: Alcatel was among the first Tier 1 vendors to commit to mobile WiMax and is
positioning the technology as complementary to GSM in emerging markets, where it sees this as
a way to leapfrog 3G. There is less emphasis on WiMax in developed markets, where the
technology is positioned either as a data overlay to 3G networks or for fixed wireless access
applications, depending on the spectrum available. The company's ASN-GW is known as the
WiMax Wireless Access Controller (WAC) and is believed to support Profile A (i.e. with split
RRM), with a view to supporting Profile C as required in the future.
UMB: As the leading supplier of CDMA equipment, Alcatel-Lucent is the vendor most committed
to UMB. As noted, from an architectural perspective, the base station functions are covered by
ongoing development of the BSR. It's not clear on which platform Alcatel-Lucent would base the
Access Gateway device, since it does not have an in-house PDSN or PDIF product.
19
Unstrung Insider Comment: Alcatel-Lucent is making an aggressive pitch for flat mobile
network architectures with its focus on the BSR concept. Its experience, particularly in distributed
RRM, should stand it in good stead relative to its competitors. The challenge, however, is to
parlay this technical leadership into a successful business, which will likely mean needing to
compromise the principles of a maximally flat network in favor of architectures aligned with mobile
operators' short-term practical requirements.
C. Alvarion Ltd.
Alvarion (Nasdaq: ALVR) is one of the largest pure-play broadband wireless access companies
worldwide and a key provider of WiMax equipment. Revenues for 2006 were $182 million, of
which WiMax accounted for $72 million.
Alvarion has been a vocal supporter of ASN Profile C and believes distributed RRM delivers
performance equivalent to that of split RRM (Profile A), with less system complexity. Moreover,
distributed RRM makes it much easier for operators to select base stations and ASN-GWs from
different suppliers, a factor that the company believes will prove a critical differentiator for WiMax
technology in coming years.
The company further believes that, while several vendors support Profile A, the market is moving
towards its approach, citing the backing of major carriers for Profile C. It notes that distributed
RRM is effective because the typical hand-over scenario would involve up to 10 or 12 sectors,
making it simple enough to manage at the base station level. Base station to base station
signaling messages are sent transparently through the ASN-GW via the R8 interface.
Alvarion says distributed RRM does not add significant cost to the base station and makes the
point that the most expensive piece of software in a WiMax base station is the scheduler used to
allocate radio resources to users. In a recent earnings call, CEO Tzvika Friedman emphasized
the company's R&D investment in this scheduler feature.
Although Alvarion supports Profile C, it has chosen to implement the ASN-GW in the base
station's transport card and does not envisage offering standalone ASN-GWs to act as
aggregation points. It will instead work with partners where such products are required; Cisco is
reportedly among them, although Alvarion is coy about the nature of the relationship. The
company foresees scenarios in which a network would use centralized ASN-GWs in urban
environments and integrated ones in less densely populated areas.
Unstrung Insider Comment: Alvarion epitomizes the challenge WiMax presents to the old order
of cellular infrastructure operators and vendors. In this sense, it's appropriate for the company to
be positioning around distributed RRM and advanced radio technology.
D. Cisco Systems Inc.

As the largest supplier of service provider routing equipment worldwide, Cisco (Nasdaq: CSCO)
is also the largest supplier of IP equipment to mobile operators. This equates to an extremely
strong presence in the IP/Multiprotocol Label Switching (MPLS) core network. It has recently
extended its presence in the transmission network with a set of RAN aggregation products. The
company was unable to arrange an interview in time for this report.
Cisco has a long-standing 3GPP and 3GPP2 mobile packet core business with many reference
accounts, but this product line appears somewhat lackluster, due to its dependence on radio
vendors as channels to market. Industry sources say Cisco has not put a huge amount of effort
into developing and selling these products, perhaps reflecting the small size of the packet core
market relative to other opportunities on which it is more focused. In the packet gateway
category, Cisco offers security gateway products based on its 7600 edge router for PDG and
PDIF applications.
20
The company is known to have ASN-GW partnerships with Navini, Alvarion, and WiNetworks. In
a positioning paper entitled Rise of the 4G Network Enabling the Internet Everywhere Experience,
Cisco discusses plans to introduce an ASN-GW for WiMax networks; sources report that the new
product would be based on the company's 7600 edge router. The paper also alludes to activities
in the SAE gateway market, although without getting into specifics, and it states that Cisco
considers the ability for the radio and IP domains to evolve separately to be a critical feature of
4G technology.
Unstrung Insider Comment: Aside from its clear strength in IP routing in the backbone and
aggregation networks, Cisco's position on mobile networks is curious. Citing unnamed sources,
Unstrung recently reported that the company has been reevaluating its WiMax options and may
be considering introduction of its own radio access product. If true, this would represent a
significant departure from its strategy of not competing against its radio access partners.
E. Ericsson AB
As the largest supplier of mobile infrastructure globally and the largest provider of GSM and
UMTS technology, Ericsson (Nasdaq: ERIC) has a dominant influence on the industry's evolution.
Like other large vendors, it enjoys a privileged position due to its contacts with a wide range of
operators. Notably, Ericsson has decided to focus its R&D efforts on the 3GPP family of
technologies, believing that it can enhance its competitive advantage by bringing its scale to bear
on fewer products lines.
Base Station Routers: Conspicuous in its absence in this emerging product sector, Ericsson
does not believe there is a business case for base stations that collapse RNC, SGSN, and GGSN
functions. Its primary reasoning is that W-CDMA networks benefit enormously from
macrodiversity, without which uplink capacity would be limited. This would create a requirement
for more sites, which add substantially to network costs. While the company concedes that it's
possible to achieve macrodiversity in the network without an RNC by combining traffic from
several base stations at the main serving base station, it maintains that the requirement of
additional transport resources would make this approach inefficient.
Direct Tunnel Architecture: Ericsson supports Direct Tunnel from the perspective of SGSN
bypass for user-plane traffic, with its first customer deployment scheduled for the third quarter of
2007. Because it uses the same hardware platform for its Node B and RNC products, it says that
combining the nodes would be technically easy. As noted, however, it does not see any value in a
collapsed RAN due to the requirement for macrodiversity and soft handover. It also refutes any
suggestion that the RNC could become a bottleneck, saying that current products have userplane throughput of 675 Mbit/s, with a roadmap for supporting up to 8 Gbit/s as packet traffic
increases and component technology advances.
Ericsson agrees that the Iub interface between Node B and the RNC is effectively proprietary and
is aware that competition among vendors serves the operators' interests. But it uses this to prove
its point with the argument that, if there were benefits to collapsed-RAN architectures, such
products would already be market successes. Therefore, it refutes the suggestion that the Iub
interface is responsible for vendor lock-in.
Additionally, Ericsson says functions such as header compression and encryption should be
centralized where possible, rather than distributed. On the encryption side, it says it is inherently
more secure to perform this function at locations not accessible to the public. For header
compression, it says that centralization delivers increased efficiency in the transport network and,
therefore, makes more sense.
E-UTRAN & EPSs: In the move to LTE/SAE, Ericsson notes that that a flatter RAN architecture is
possible because macrodiversity is not specified (OFDMA uses fractional frequency re-use
between adjacent sectors to manage interference), and therefore there is no requirement for an
21
RNC. The absence of an RNC forces functions such as header compression and encryption to
move elsewhere in the architecture, with the base station being the most obvious home. In the
SAE core, Ericsson says it is considering several implementation options and that there is still
uncertainty regarding how operators will deploy networks, as well as how requirements will evolve
over time.
Options under consideration include supporting the MME function in the SGSN alongside 2G and
3G technology. Ericsson notes that most of its customer now use combined 2G/3G SGSNs,
making the addition of MME straightforward and logical. The company is also evaluating the
possibility of centralizing the control plane functions of the 3G SGSN and MME, with potentially
more distributed user-plane gateways, but says this is a longer-term project.
Ericsson plans to support the SAE gateway on its existing GGSN platform with the expectation
that operators will want to run both applications simultaneously. It prefers the single-SAE gateway
option under discussion in 3GPP, but would offer the dual-gateway approach if required.
Unstrung Insider Comment: Ericsson's rejection of the collapsed 3G RAN concept sets it apart
from other Tier 1 vendors and many Tier 2 suppliers. It looks set to stick with the position and
says it has invested substantial effort making its case with network operators, reporting that this
campaign has largely met with success.
The challenges for Ericsson in the packet core will likely be guarding against complacency and
investing enough in product development. Because this market is not hugely lucrative, the
temptation will be to invest in R&D on an opportunity-led basis and not make strategic
development decisions early enough. That said, the company's plans are simplified by not
supporting multiple standards roadmaps.
F. Huawei Technologies Co. Ltd.

Huawei does not have a large installed base in the mobile packet core, having focused instead on
winning market share in the more lucrative 3G RAN and mobile softswitch segments. Although
not interviewed directly for this report, previous briefings indicated development of a common
packet gateway platform for use across different wireless access network, including 3G and
WiMax. The company is not seen as leading the push to flat architectures in the 3G networks; for
instance, its femtocell project is very much oriented around the RNC integration model.
In the WiMax sphere, Huawei's Wireless Access Service Node, based on its fifth-generation core
router platform, is positioned as a centralized ASN-GW that supports up to 600,000 simultaneous
active connections and to scales to 5 Gbit/s per rack. It shipped its first mobile WiMax products at
the end of 2006. The company reports having more than 600 engineers working on WiMax R&D
alongside LTE/SAE in a common development program. The WiMax base station and ASN-GW
platforms will be used for Enhanced Node B and ASN-GW applications, respectively.
Unstrung Insider Comment: The impression given is that Huawei is customer-led with respect
to network architecture options and is prepared to invest man-hours in delivering products to
support a carrier's chosen architecture. However, with the company now an established global
player, it should look to provide more leadership on next-generation mobile network evolution.
Although clearly very active in the standards bodies, it does not always articulate it overall
architectural visions very clearly.
G. Juniper Networks Inc.

Juniper (Nasdaq: JNPR) doesn't compete directly in the classic wireless infrastructure market, but
as the leading supplier of IP networking equipment to mobile operators, it clearly has significant
influence and a vested interest in the move to flat network architectures. Juniper's primary areas
of interest are the IP/MPLS core, RAN aggregation, security, and policy control. The company
has also talked in the past about developing products for a "unified mobile edge" rather than
22
jumping into the wireless market with specific application nodes, but it is not yet ready to share
more detail on this initiative.
With respect to the move to all-IP mobile networks, Juniper is looking to extend the capabilities of
MPLS from the core to the network edge and envisions a role for MPLS as far out as the cell
sites. It notes, however, that MPLS will be much less dynamic in the RAN, since operators are
uncomfortable with the concept of dynamic routing in this part of the network. Instead, it believes
MPLS domains will be subdivided into manageable sizes, each separated by security gateways.
Juniper also believes it can play a role in policy management and enforcement through its AAA
and policy management applications. It holds that, given the importance of end-to-end QOS,
admission control should be performed at the first point of entry into the network the base
station. As it doesn't sell RAN equipment, the company is focused on a partnership strategy that
would see it working with specialist mobile vendors.
Unstrung Insider Comment: As things stand, Juniper does good business in the mobile market.
The challenge will be the move up the stack from routing to participating in specific mobility
applications. Its partnership strategy seems a logical way to approach this but leaves it a little
exposed to the whims of the large RAN vendors.
H. Motorola Inc.
Although a player in 3G and specifically CDMA, Motorola (NYSE: MOT) has decided to put its
R&D resources behind mobile WiMax, which it believes will emerge as the most cost-effective
next-generation wireless standard. The company is understood to be one of very few backers of
Profile B in the WiMax ASN, with all functionality integrated into the base station. However, it also
offers a standalone ASN product called the APC 1000 Access Point Controller, about which
details have not yet been made public.
In the 3GPP market, Motorola has pitched the idea of collapsed-base station architecture with its
AXPT 3G access point, a picocell product intended for enterprise and hotspot use. However, very
little has been heard of the product in the market, and sources say it has been quietly shelved.
The company did not arrange for a representative to be interviewed in time for this report.
Unstrung Insider Comment: Beyond stating that it is behind WiMax, Motorola's next-generation
wireless infrastructure strategy is not clear, especially as relates to 3GPP and 3GPP2 networks
I. NEC Corp.
NEC (Nasdaq: NIPNY) is a large supplier of 3GPP RAN and packet core equipment and could be
said to have pioneered to the move to semi-flat architectures with the 2004 launch of its
integrated SGSN/GGSN product. It has since rolled out PDG (security gateway) and RNC
applications on the same ATCA-based platform.
Although the company was not interviewed for this report, there is no indication that NEC is
looking at RNC bypass in the macro network. Instead, it is positioning its new iRNC products for
use in a distributed IP RAN architecture. It is, however, an aggressive adopter of the flat
architecture in the 3G femtocell market, with plans to integrate femtocells from partner Ubiquisys
via the I-WLAN architecture using its in-house PDG.
Documentation on NEC's mobile WiMax portfolio shows that it supports RRM on the ATCA-based
ASN-GW ( la Profile A), rather than distributed to the base stations.
Unstrung Insider Comment: NEC was an early adopter of the common platform principle for
multiple applications and is rarely found lacking technically, especially in W-CDMA. What's
interesting is how the company is repositioning itself more aggressively in Europe and other
international territories since Siemens (its 3G partner) undertook its joint venture with Nokia.
23
J. Nokia Siemens Networks B.V.

As one of the big three wireless equipment suppliers, Nokia Siemens a joint venture of Nokia
Corp. (NYSE: NOK) and Siemens AG (NYSE: SI; Frankfurt: SIE) wields enormous influence on
the direction of industry development and takes a leadership role in defining next-generation
mobile network architectures.
Base Station Routers: Nokia Siemens is not active in this product category, although it may
adopt the collapsed-RAN architecture for femtocell deployment in the future. The company's
femtocell plans are currently under wraps, but some public confirmation of its anticipated
participation in this market is expected later in 2007.
Direct Tunnel Architecture: From a practical perspective, this is Nokia Siemens's strategic
response to the demand for flatter networks, with today's initiatives aligned closely with its longterm product roadmap. The company has been pushing for the inclusion of the Direct Tunnel
approach into 3GPP for three years and was delighted to see it accepted in 2006. It prefers the
flattest Direct Tunnel option, with the RNC integrated in the base station and the SGSN retained
only for limited control-plane functions.
Nokia Siemens believes it contributed significantly to efforts by 3GPP members to integrate
features such as encryption and header compression in the base station, and notes that this
architecture has been carried forward in the LTE/SAE specification work. Similarly, it believes the
standardization of carrier sharing (classic HSPA and Evolved HSPA on the same transceiver) and
handover between evolved and classic networks will prove very important to the early introduction
of flat architectures.
As noted, the Direct Tunnel approach is closely linked to the I-HSPA initiative that was
announced by Nokia in 2005. In terms of practical implementation, Nokia Siemens has ported the
RNC software to its Flexi BTS Node Bs for new deployments and developed an I-HSPA adapter
that can be retrofitted to Flexi BTS Node Bs already in the field. In a second release, its older
Ultrasite Node Bs and Node Bs from Siemens will also be supported via a software upgrade to
the transport cards.
Nokia Siemens believes I-HSPA will be deployed by greenfield operators looking to build datacentric mobile networks and as an overlay to existing 3G RANs. The idea is that operators could
use I-HSPA to overlay the current RAN in areas where there is high demand for data services.
The most logical time to take this step would be when an operator adds capacity to its sites by
deploying new frequency carriers. Importantly, Nokia Siemens says this overlay will break the link
between increasing traffic and the requirement to pay vendors capacity-based fees to upgrade
RNCs and SGSNs.
The company will begin second-phase trials in the second half of 2007, with general availability
expected in early 2008. It has tested with one third-party GGSN vendor to date.
E-UTRAN & EPSs: It is notable that Nokia Siemens has combined its I-HSPA architecture and
LTE/SAE R&D teams. The vendor believes the initiatives have high-level similarities in terms of
key functions' locations in the network (although, obviously, the radio interface is different). The
company cites the integration of encryption and header compression in the base station as an
example and notes that, while technically different, adding the Iur interface to a Node B is roughly
equivalent to the X2 interface specified in the E-UTRAN.
On the core network side, Nokia Siemens plans to support the SAE Gateway on its Integrated
Services Node platform, which is currently used for the GGSN application. It envisages both the
SAE and GGSN applications running simultaneously on the same hardware and reports that its
LTE/SAE schedule is line with the rest of the industry's.
24
WiMax ASN: Nokia-Siemens's jump into the mobile WiMax market was a surprise from some
perspectives, given that both Nokia and Siemens had previously expressed reservations about
the technology's value. Siemens had a fixed WiMax division, but saw no major reason to go
mobile; the joint venture will go forward with the Nokia-originated mobile WiMax product.
On the base station side, this makes sense enough; the Flexi BTS was designed as a multistandard radio platform. The ASN-GW, however, appears to be a new effort not based on the
same software or hardware platforms as those being used for 3GPP products. Very little
information has been released on the ASN-GW, but a public presentation shows it sitting in front
of the Integrated Services Node used for 3GPP gateways. (See slide 18 of the Nokia PowerPoint
presentation "WiMax, Personal Mobile Broadband for the Masses.") The company prefers ASN
Profile C and the distributed RRM approach.
Commercial availability of both the WiMax base station and the ASN-GW is scheduled to coincide
with the launch of Sprint's WiMax service in early 2008.
Unstrung Insider Comment: Presenting a common front on mobile network architecture
evolution is one clear success of the Nokia Siemens Networks merger. In this respect, the new
company has a strong and well-communicated strategy. Perhaps the best example is the
consolidation of the two companies' LTE and SAE R&D teams to complement work on nearerterm I-HSPA products.
K. Nortel Networks Corp.

Since its exit from the UMTS access market in 2006, Nortel Networks (NYSE: NT) has thrown its
weight behind WiMax and OFDMA technologies in general. It also intends to support both the
3GPP and 3GPP2 core network roadmaps and has ambitions in the next-generation radio access
market specified as UMB or LTE, evidenced by its support for OFDMA radio technologies.
Nortel sees the core network evolving to an all-IP infrastructure with a common service edge that
offers consistent capabilities and features, such as policy control and enforcement, across
wireless access networks. The intent is to deploy this service edge in a manner consistent with an
IMS control plane.
From a product strategy perspective, Nortel is developing a common access gateway platform
based on ATCA hardware that is physically scaleable enough to support various cost and traffic
profiles. This platform will aggregate the functions of multiple technologies and standards,
allowing operators to support different RANs simultaneously on the same equipment.
On the WiMax side, Nortel's ASG 5000 ASN-GW will be available later this year in line with the
launch of its wider WiMax solution. The company supports the split-RRM model (Profile A). On
the 3GPP2 side, Nortel will look to reuse operators existing RNC and PDSN assets while also
positioning its new hardware platform, which it says will be upgradeable in software to support 4G
access technologies. In the 3GPP2 core, Nortel has a longstanding relationship with Starent that
appears to have hit turbulence recently as discussed in the Starent profile below. On the 3GPP
side, Nortel plans to develop its existing SGSN and GGSN software and interfaces to provide
MME nodes and SAE gateways on the new hardware platform.
Unstrung Insider Comment: Nortel's adoption of a common ATCA-based access gateway
platform for multiple applications and standards is exactly in line with market requirements. The
challenge will be timing the product introduction to capture the bulk of the packet core refresh
business being considered across Western Europe. The company has a good share of this
market with its legacy Shasta-based GGSN product and must ensure that its erstwhile partner
Starent doesn't take advantage of any delay in product development. On the positive side, its
competitors among the big 3G RAN vendors don't have fully convincing common-platform stories
for the core.
25
L. Samsung Electronics Co. Ltd.

Although a longtime player in 3G, it is clear Samsung (Korea: SEC) has bet its future in mobile
infrastructure on mobile WiMax. Indeed, it could be said to have laid the foundation for mobile
WiMax with its participation in the Korean wireless broadband (WiBro) initiative. With that
deployment, the company introduced its Access Control Router product, which is roughly
equivalent to an ASN-GW but looks to have conceptual roots in the traditional RNC model used in
3G networks. For example, it currently supports split RRM (Profile A) and places functionality
such as header compression in the ASN-GW.
Samsung also plans to introduce a product capable of distributed RRM (Profile C) in mid 2007,
indicating a change of mind over the course of its WiBro/WiMax development initiative. Sources
outside the company say large carriers have encouraged this; the company itself was not
interviewed for this report.
In the 3G packet core markets, Samsung has a partnership with Starent for PDSNs and GGSNs.
In the femtocell space, it is pitching its CDMA Ubicell product with integrated RNC.
Unstrung Insider Comment: Samsung is going all-in on mobile WiMax and, despite its legacy
position on CDMA, this initiative is the focus of its next-gen wireless development.
M. Starent Networks Corp.

Since its creation in 2000, Starent (Nasdaq: STAR) has become the leading independent provider
of mobile packet core systems; it has around 400 employees, of whom over 200 are located in
India. In March 2007, the company filed for an IPO, which it completed in June, raising a total of
$108 million. It had raised approximately $100 million in venture funding to that point.
In 2006, Starent reported revenues of $93 million and a net profit of $3.6 million. More than 40
percent of revenues, of which more than half was tied to a single mobile operator, were
generated through an OEM relationship with Nortel Networks. A further 30 percent of revenues
came from Verizon Wireless. The company spent $26 million on R&D in fiscal 2006. It terminated
its OEM relationships with Nortel Networks in December 2006.
As a specialist packet core vendor, Starent takes a different view of the market than do larger
end-to-end system vendors. Despite generating the majority of its revenues from CDMA EV-DO
operators, the company's core value proposition is essentially radio-agnostic. Its plan is to
support multiple access networks off a common hardware and software platform. Current
supported applications include GGSN, SGSN, and PDG for 3GPP networks; Home Agent, PDSN,
and PDIF for 3GPP2 networks; and an ASN-GW for WiMax networks. Notably, the company's
products are able to support multiple applications from different standards groups simultaneously
on the same piece of equipment, a factor that could prove to be a decisive differentiator.
Starent currently offers two products the ST16 Intelligent Mobile Gateway and the newer,
higher-capacity ST40 Gateway. Both products are based on a 16-slot chassis with custom
operating system, backplane, and line cards that provide pooled processing for applications
configured on the equipment. The idea is that each "Packet Services Card" can execute all the
features required by a particular application, with load distributed across the box as required.
Starent believes this to be another differentiator, as competitors' products often require
specialized line cards for different functions.
Starent particularly touts its ability to support in-line services without increases in latency. For
example, the company uses deep packet inspection technology for policy enforcement and
content filtering, in addition to offering services such as firewalls, virus scanning, IP address
management, and cross-network mobility. It argues that its approach is far superior to supporting
best-of-breed applications on separate servers, since the streamlining reduces latency and
simplifies management.
26
Citing a quiet period related to its recent IPO, Starent is coy about the specific applications it
intends to support in the future. Nevertheless, support for LTE/SAE and UMB appears an obvious
next step.
As an IP equipment vendor, Starent is obviously positive about moves toward flat architectures,
yet it positions itself as a supporter of the "semi-flat" network, in which packet gateways provide
the operator with invaluable features beyond simple base station aggregation functions. As
compared to fully-flat architectures, Starent maintains that superior Layer 3 mobility is a key
benefit of dedicated gateways.
Unstrung Insider Comment: Starent has the industry's most convincing packet core evolution
story and has clearly benefited from focusing on this specific market segment an advantage not
enjoyed by its larger competitors. It is the only supplier that supports such a wide range of
applications on a common platform today, and one of few suppliers to articulate a persuasive
common platform strategy. While it is the dominant provider of packet core equipment for the
3GPP2 community, Starent's move into the much larger 3GPP market will be far more
challenging and more closely watched.
27
IX.
Conclusion
The move to IP applications is driving a rethink of mobile network architectures. Flat networks are
characterized by fewer network elements, lower latency, greater flexibility, and lower operational
cost. There is remarkable agreement as to what networks should look like in future: base station
router products interconnected by IP/Ethernet, deployed in a flat user-plane architecture, with
services provisioned and managed by an IMS control plane. Getting there is the challenge.
Key to emerging flat mobile network architectures is advanced base station equipment that
integrates functions such as radio control, header compression, encryption, call admission control
and policy enforcement. This model of distributed user-plane functionality marks a departure from
today's hierarchical networks. Questions remain, however, about how to implement RRM and
where it should be located in the network. Distributed RRM (in the base station) enables a more
open ecosystem and flatter network architecture, but there are concerns as to its potential
performance impact and practicalities.
Operators are also looking to decouple radio access from the core network in an effort to increase
flexibility, creating demand for "access gateway" products capable of supporting multiple access
technologies simultaneously. Most vendors are moving toward common hardware and software
platforms, but few currently offer products that meet this challenge.
The WiMax network architecture has pressured the 3G community to respond to the flat network
concept, and many of the ideas implemented in WiMax have since found their way into the
standards evolution at the 3GPP and 3GPP2. As WiMax matures, however, vendors and
operators are introducing complex solutions to enhance performance and increase differentiation.
On the 3GPP track, the most practical and near-term implementation of the flat network principle
is the Direct Tunnel architecture. However, vendors disagree as to whether this approach should
be restricted to SGSN bypass or extended to collapse RNC functionality into the base station.
base station router products offer promise for UMTS and CDMA 2000 3G networks, but they face
practical implementation challenges. This is likely to hinder near-term uptake in the macro
network; the most obvious use for such products is in femto-, pico-, and microcell applications.
As noted, although there is agreement on the long-term vision for flat networks, there are multiple
ways to achieve this, depending on an operator's starting position and migration path. Figure 16
summarizes the positioning of major vendors relative to their options for implementing flat
network architectures. It is interesting to note that some vendors for example Alcatel-Lucent,
Motorola, Nortel, and ZTE are spread across a wide range of technologies; while others
including Nokia Siemens and Huawei are focused on two primary technologies. There are also
companies that prefer to specialize, Ericsson being the best example.
Figure 16: Summary of Vendor Positioning on Flat Mobile Network Architectures
Supplier
Base Station
Router
3GPP Direct
Tunnel
WiMax
LTE
UMB
Airvana
Planned for
CDMA & 3G
femtocells
Not active
Not active
No information
No information
AlcatelLucent
Leading
vendor; BSR
targeted at
3G & 4G
Likely to
support RNC
& SGSN
bypass
Centralized
ASN-GW;
Profile A
Aligned
with WiMax
development
Lead vendor
behind this
initiative
28
Supplier
Base Station
Router
3GPP Direct
Tunnel
WiMax
LTE
UMB
Alvarion
Integrates
ASN-GW with
WiMax base
stations
Not active
Distributed
ASN-GW;
vocal backer
of Profile C
Not active
Not active
Cisco
Supports in
principle; may
provide tech
components
No information
Centralized
ASN-GW;
Profile C
May provide
specific tech
components
via partners
May provide
specific tech
components
via partners
Ericsson
Not active
Supports
SGSN bypass;
retains RNC
Not active
Common core
platform for
3G & LTE
Not active
Huawei
No information
No information
Centralized
ASN-GW; no
info on profile
Aligned
with WiMax
development
No information
Juniper
Supports in
principle; may
provide tech
components
Not active
May provide
specific tech
components
via partners
May provide
specific tech
components
via partners
May provide
specific tech
components
via partners
Motorola
Enterprise
picocell for
UMTS
Not active
Distributed
ASN-GW;
Profile B
No information
Expected to
participate
NEC
For 3G
femtocells
First with
integrated
SGSN/GGSN
Centralized
ASN-GW;
Profile A
No information
Not active
Nokia
Siemens
Via its I-HSPA

initiative
Lead advocate
via I-HSPA
Centralized
ASN-GW;
Profile C
Aligned
with I-HSPA
initiatives
Not active
Nortel
No information
Not active
Centralized
ASN-GW;
Profile A
Aligned
with WiMax
development
Aligned
with WiMax
development
Samsung
For 3G
femtocells
Not active
Centralized
ASN-GW;
profiles A & C
No information
No information
Starent
Not active
Integrated
SGSN/GGSN
Not active
Likely to offer
LTE core
Likely to offer
UMB core
Research Analyst: Gabriel Brown (brown@unstrung.com)

Support: www.unstrung.com/insider (insider@unstrung.com)
29
SUBSCRIBER LICENSE AGREEMENT

Any Unstrung Insider report ("Report") and the information therein are the property of or licensed
to Light Reading Inc. ("Light Reading") and permission to use the same is granted to annual or
single-report subscribers ("Subscribers") under the terms of this Subscriber License Agreement
("Agreement") which may be amended from time to time without notice. When requesting a
Report, Subscriber acknowledges that it is bound by the terms and conditions of this Agreement
and any amendments thereto. Light Reading therefore recommends that you review this page for
amendments to this Agreement prior to requesting any additional Reports.
SUBSCRIPTION RENEWAL
For convenience, subscriptions purchased by credit card will be auto-renewed at the end of the
subscription term. Prior to that auto-renewal you will receive a notice that will include instructions
for updating your subscription or payment information and for opting out of the auto-renewal
process.
OWNERSHIP RIGHTS
All Reports are owned by Light Reading and protected by United States Copyright and
international copyright/intellectual property laws under applicable treaties and/or conventions.
Subscriber agrees not to export any Report into a country that does not have
copyright/intellectual property laws that will protect Light Reading's rights therein.
GRANT OF LICENSE RIGHTS
Light Reading hereby grants Subscriber a personal, non-exclusive, non-refundable, nontransferable license to use the Report for research purposes only pursuant to the terms and
conditions of this Agreement. Light Reading retains exclusive and sole ownership of each Report
disseminated under this Agreement. Subscriber agrees not to permit any unauthorized use,
reproduction, distribution, publication or electronic transmission of any Report or the
information/forecasts therein without the express written permission of Light Reading.
Subscribers purchasing site licenses may make a Report available to other persons from their
organization at the specific physical site covered by the agreement, but are prohibited from
distributing the report to people outside the organization, or to other sites within the organization.
Enterprise Level Subscribers, however, may make a Report available for access on computer
intranets or closed computer systems for internal use under their service agreements with Light
Reading.
DISCLAIMER OF WARRANTY AND LIABILITY
Light Reading has used its best efforts in collecting and preparing each Report.
LIGHT READING, ITS EMPLOYEES, AFFILIATES, AGENTS, AND LICENSORS DO NOT
WARRANT THE ACCURACY, COMPLETENESS, CURRENTNESS, NONINFRINGEMENT,
MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE OF ANY REPORTS
COVERED BY THIS AGREEMENT. LIGHT READING, ITS EMPLOYEES, AFFILIATES,
AGENTS, OR LICENSORS SHALL NOT BE LIABLE TO SUBSCRIBER OR ANY THIRD PARTY
FOR LOSSES OR INJURY CAUSED IN WHOLE OR PART BY OUR NEGLIGENCE OR
CONTINGENCIES BEYOND LIGHT READING'S CONTROL IN COMPILING, PREPARING OR
DISSEMINATING ANY REPORT OR FOR ANY DECISION MADE OR ACTION TAKEN BY
SUBSCRIBER OR ANY THIRD PARTY IN RELIANCE ON SUCH INFORMATION OR FOR ANY
CONSEQUENTIAL, SPECIAL, INDIRECT OR SIMILAR DAMAGES, EVEN IF LIGHT READING
WAS ADVISED OF THE POSSIBILITY OF THE SAME. SUBSCRIBER AGREES THAT THE
LIABILITY OF LIGHT READING, ITS EMPLOYEES, AFFILIATES, AGENTS AND LICENSORS,
IF ANY, ARISING OUT OF ANY KIND OF LEGAL CLAIM (WHETHER IN CONTRACT, TORT
OR OTHERWISE) IN CONNECTION WITH ITS GOODS/SERVICES UNDER THIS
AGREEMENT SHALL NOT EXCEED THE AMOUNT YOU PAID TO LIGHT READING FOR USE
OF THE REPORT IN QUESTION.
30

Flat IP Architectures For Mobile Networks - Unstrung Insider Report

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Flat IP Architectures For Mobile Networks - Unstrung Insider Report

Hochgeladen von

Copyright:

Verfügbare Formate

VOL. 6, NO.

Flat IP Architectures for

Introduction: Flat Is Back

Evolving Today's Networks

Base Station Routers

UMTS Enhancements Direct Tunnel Architecture

WiMax Network Architecture

3GPP System Architecture Evolution

CDMA Enhancements Ultra Mobile Broadband

VIII. Company Profiles

Introduction: Flat Is Back

Enhancements to existing 3GPP specifications: intended to support High Speed Packet

Worldwide Interoperability for Microwave Access (WiMax) access services network

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Evolving Today's Networks

A. Drawbacks of Existing Network Architectures

Source: Unstrung Insider

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Such a scenario has several drawbacks:

Source: Unstrung Insider

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Source: Unstrung Insider

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Base Station Routers

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

UMTS Enhancements Direct Tunnel Architecture

A. Background to 3GPP's Direct Tunnel Approach

Direct Tunnel with retained RNC

Direct Tunnel with RNC control plane

Direct Tunnel with collapsed RNC

B. Direct Tunnel With Retained RNC

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Figure 5: Variation on Direct Tunnel Approach in 3GPP RNC Retained

Source: 3GPP TR 25.999 (March 2007)

C. Direct Tunnel With RNC Control Plane

Source: 3GPP TR 25.999 (March 2007)

D. Direct Tunnel With Collapsed RNC

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Figure 7: Variation on Direct Tunnel Approach in 3GPP Collapsed RNC

Source: 3GPP TR 25.999 (March 2007)

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

WiMax Network Architecture

Source: Intel Corp.

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Source: Nokia Siemens Networks

Intra-ASN: between base stations connected to the same ASN-GW

Inter-ASN: between base stations connected to different ASN-GWs

Inter-CSN: between ASNs connected to different core networks

Profile B: ASN integrated in the base station

Profile C: RRM integrated in the base station, with a separate ASN-GW

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Supports soft handover

Simpler, lower cost for

Easier to use different

Soft handover is difficult;

Source: Unstrung Insider

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

3GPP System Architecture Evolution

Source: Source: 3GPP (TS 36.300, Draft, April 2007)

B. Evolved Packet System

UNSTRUNG INSIDER | VOL. 6, NO. 6, JUNE 2007 | LIGHT READING

Source: 3GPP (TS 23.401, Draft, May 2007)