3GPP IWLAN – a Closer Look at One of Two Paths to FMC

By Kevin Mitchell
Service providers are implementing fixed-mobile convergence (FMC) architectures that aim to deliver access to IP data and interactive communication services over wireline, wireless and blended broadband connections. FMC embodies various access and core technologies and the drive for FMC is mandating evolution in network architectures and introducing new functions and network elements. The evolution of FMC can be roughly drawn along a path from tunneled GSM signaling over WiFi to full blown SIP in an IP RAN with a focus today on delivering services over a combination of nearly ubiquitous WiFi with mobile radio access networks. In order for FMC to be successful for operators, the cost of implementing the various architectures will be determined in part by how the new functions can deployed.

Two 3GPP FMC Architectures

The 3rd Generation Partnership Project (3GPP) defines two specific FMC architectures regarding GSM networks:

• Unlicensed mobile access (UMA) – uses GSM signaling for voice services over the circuit-switched radio access network GERAN (GSM EDGE Radio Access Network) as well as the IP WiFi access network

• Interworking-wireless LAN (IWLAN) – uses GSM signaling for voice services over circuit-switched UTRAN (UMTS Terrestrial Radio Access Network) access but SIP signaling for voice services over WiFi access

Both approaches require dual mode handsets that support 2 radio interfaces, of which there are a growing number on the market today. Femtocells can also take advantage of UMA or SIP for the femto base stations at the premise. In all cases the WiFi access network makes use of fixed network backhaul (such as DSL, WiMAX or cable) using the general Internet (e.g., a mobile operator using the Internet connection of a cable provider to connect to dual-mode handsets) or a managed IP network (e.g., a fixed-mobile operator that uses its private IP network), with implications for security and call quality.

For service providers both FMC approaches share advantages, such as improved indoor coverage and reduced backhaul costs, and drawbacks like limited handset selection today and billing complexities. As the signaling protocols differ between the approaches, there are some benefits unique to each FMC architecture.

UMA enables service transparency and seamless roaming between GSM and WiFi as the mobile core network remains largely unchanged using existing mobile switching centers (MSCs). The back office and OSS do not require any changes, making rollout easier. Initially UMA was solely specified for 2G /2.5G RANs and not the growing base of 3G voice and packet data services, but there has been some work to support UTRAN environments.

3GPP IWLAN is future-oriented and as such is specified only for 3G RANs and handsets and requires SIP and core IMS equipment investments and depends on an embryonic network function—the voice continuity server (VCC)—to enable roaming from and to WiFi and GSM access networks. However, IWLAN primarily addresses the ability to handle operator roaming; i.e. how to handle visited networks which is key for widespread adoption. As this FMC architecture is tightly integrated with IMS, IWLAN enables access to new services delivered by the IMS core infrastructure.

Both UMA and IWLAN approaches require a new network function: a security gateway that authenticates devices and terminates the IPsec tunnels from the handsets when on a WiFi network. This article focuses on deployment considerations for IWLAN and the placement of the tunnel termination gateway (TTG) function.

3GPP IWLAN Architecture Overview

In 3GPP IWLAN a dual-mode handset (or user equipment, UE), supporting both SIP and GSM signaling, accesses the UTRAN network or WiFi network depending on the radio signal strength. Both voice and data services are delivered via the same access networks to which the UE is connected at that time.

When services are delivered via the UTRAN, voice is circuit-switched, using traditional MSCs. Packet data services are delivered via the IP portion of the UTRAN and are served by the existing SGSNs (Serving GPRS Support Node) and GGSNs (Gateway (News - Alert) GPRS Support Node). (See Chart 1.)

When the UE is connected to a WiFi access network all voice and data services are encapsulated in an IPsec tunnel fixed network backhaul network such as DSL. In this scenario, SIP-based interactive communication services are controlled by a session border controller (SBC) providing the 3GPP-defined P-CSCF and C-BGF functions in the IMS network. IP data services and IP address management are controlled by the SGSN and GGSN. A VCC application server enables handover and roaming between the circuit-switch GSM voice and IP-based SIP voice. The new function that’s defined in IWLAN is the TTG which authenticates devices (via AAA queries), decrypts sessions originating from mobile handsets, allocates IP addresses and protects the layer 3 and key exchange infrastructure from denial-of-service attacks.

One of the main decisions for service providers revolves around the degree of separation or integration of the TTG function with established SBCs and GGSNs.

IWLAN Architecture Deployment Considerations

3GPP defines functional elements, not specific products. The standards groups do not dictate or provide guidance on the elements that can be combined into a single product, but does define interfaces between elements in the event they are distinct products. However, multiple functional elements could be integrated into a single product. This integration can have an effect on the scalability, manageability and cost of the individual elements and the overall fixed mobile convergence network.

As GGSNs process and route IP data packets, the addition of IPsec termination can be seen as a natural fit. The TTG function can be logical or physically combined with a GGSN or subset of GGSN functionality and are referred to as a packet data gateway (PDG). Logical PDGs will exist with standalone TTGs with interfaces to distinct GGSN. The TTG can also be combined with SBC components (P-CSCF and C-BGF) as the SBC is the first signaling hop in a service provider network and the security element for IP interactive communication services.

Creating a leading GGSN or SBC is very difficult due to the intensive software development time and costs, interoperability, high availability and hardware design for scale and performance. They are also significantly different products in that they each inspect and control a specific type of network traffic. They can also be deployed in different parts of the network (SBCs on the broadband access edge or edge of IMS core and GGSNs in mobile core or the edge of walled garden server farm). Due to the stark differences, the requirements for processing the different traffic types and the challenges in building SBCs and GGSNs, the likelihood of these functions being combined into a single element is remote. Adding the TTG function to either platform is comparatively easy and relatively inexpensive.

Operators should be wary of the “god box” approach to fixed-mobile convergence with vendors purporting to solve all challenges and deliver all requisite functional elements in a single network element. Various god boxes have been proposed in telecommunications and have fallen short in the promise to deliver best-of-breed functionality across many functional domains along with the scale, performance and high availability requirements while being cheaper to buy and operate than dedicated network elements.

Pragmatic network design and selection for TTG must take into account many factors, including:

• Session composition – number of SIP sessions, non-SIP walled garden data and Internet access

• System throughput – aggregate system bandwidth (Gbps) for handling VoIP and data

• Capacity – number of IPsec tunnels and sessions and whether they are appropriately in proportion to each other

• Capital expenditures – cost per IPsec tunnel

• Integration tradeoff – degree of impact in terms of scale or performance on the core functions with which the TTG is integrated

• Operational expenditures – number of network elements required, rack space and power draw

• Service evolution – FMC architecture used today and how the elements may work in the evolution path from UMA with GSM signaling to TTG with SIP signaling to 4G VoIP over licensed spectrum

• Location dependent usage – mobile and premise-based usage will be different; for instance, consumers will likely watch IPTV and movies on big screen wall-mounted TVs, not the small screens on handheld mobile devices

• Physical location – the location of pre-existing and planned network elements may dictate the feasibility and possibility of TTG integration

• Degree of integration – functions could occupy the same physical rack or chassis, yet involve external interfaces or be tightly coupled as a single functional element

• Access network ownership – integrated mobile and wireline operators can make different decisions than mobile-only or wireline-only providers based on the access networks it controls and owns

As noted in the above list, one of the top concerns that should drive the decision for placement of TTG functions is the number of sessions (and revenue) controlled by each device — SBCs for SIP-based interactive communications and GGSNs for walled garden packet data services. In most cases, the Internet access and web-based services accessed from a mobile device do not need to be encrypted and may not be backhauled to the operator’s mobile core to the GGSN. Instead, the user can access the Internet via the local broadband connection (DSL, cable, etc.) that is backhauling the traffic from the femtocell or dual-mode handset.

In an IMS environment the SIP-based services — those controlled by session border controllers — include VoIP, text and multimedia messaging and video sharing. IMS also brings the promise of additional service types using SIP such as gaming, multimedia collaboration and other services. The packet data services controlled by the GGSN today include SMS, ring tones, walled garden services that may include videos, news, sports, games and ringtones. On the GSM RAN, mobile packet core elements also tend to handle and create billing records for Internet access to Google searches, YouTube (News - Alert) videos, online banking etc.

Research firm Frost & Sullivan reports that North American mobile operator revenue is vastly voice and predicts it will remain that way. Frost & Sullivan reports that in 2006, mobile voice revenues in North America totaled US$131 billion dollars. The packet data revenue, while large in its own right at $13 billion, is only 8% of the total mobile revenues of $144 billion. That figure also includes SMS and MMS revenues, services that are transitioning to SIP signaling services in IMS networks. (See Graph 1.)

Compared to revenues, the number of sessions (i.e., calls) more heavily favors voice. Consider the average mobile subscriber’s phone bill: the ratio of the number of calls made and received versus the number of messages sent or ringtones downloaded is easily in the range of 100-500 to 1. As the revenue and sessions are vastly SIP-signaled services, this heavily favors the integration of the TTG function with the first SIP device on the service provider edge. As there is a long list of factors that should drive this decision, there will be cases where GGSN integration may be favored or the only possible choice.

Summary

Today, there are specific network elements tied to enabling voice and data services in mobile networks—session border controllers for VoIP and mobile packet core elements (i.e., SGSNs and GGSNs) for data services. The FMC architectures, as defined by 3GPP, introduce a new functional element centered around mobility management, subscriber authentication and secure and encrypted access between a mobile device and the mobile core network serving the subscriber. 3GPP IWLAN operators are facing a decision in 2008 on how to deploy this new function, the tunnel termination gateway: as a standalone element or integrated in a SBC or GGSN. If integration is preferred, given that fixed and mobile voice communications are moving to SIP and more individual sessions and revenue-generating traffic will traverse an SBC rather than a GGSN, the TTG function is likely to be integrated with the session border controller.

Kevin Mitchell (News - Alert) is Director, Solutions Marketing, for Acme Packet (www.acmepacket.com). Reach him at kmitchell@acmepacket.com.

Other Network Elements

Session Border Controllers (SBCs): provide critical control functions to deliver high quality interactive communications across IP network borders; the SBC controls and shapes any real-time, interactive voice, video or multimedia communication using IP session-layer signaling protocols such as SIP and data services that are not set-up via signaling protocols are not processed by a SBC.

Serving GPRS Support Nodes (SGSN) and GPRS Gateway Support Nodes (GGSNs): essentially, edge and core routers for GSM networks that handle IP address allocation, layer 3 routing, traffic shaping, mobility management for packet data applications; there is no SIP session control or awareness.

IMS Magazine Table of Contents