The traditional GSM/UMTS cellular operators have mandated that femtocells connect directly to existing core network elements using a standard Iu interface. This allows seamless handover between femtocells and existing 2G/3G cellular networks, with the same services, phone numbers and ease of use that we are all used to. It is also compatible with existing mobile phones - no changes or upgrades are required.
However, several femtocell vendors have been championing the use of IP based call handling protocols instead of the current well-trodden and adopted ones used in both fixed and mobile networks today. Recent announcements from Airvana suggest that this approach is finding favour amongst CDMA operators, with both Hitachi and Alcatel-Lucent announcing agreements to develop joint CDMA/IMS solutions. This appears to be specifically for the Japanese market, one of the countries with high CDMA technology adoption. Softbank, a Japanese UMTS operator (formerly Vodafone KK) has also been reported to favour using a SIP/IMS architecture .
With 4G adopting the LTE radio interface and SAE network architecture which is based on IMS, can femtocells be used to set a roadmap for IMS trial and adoption?
Some background about mobile network switching protocols
Traditional mobile networks use a derivative of the CCITT Signalling System Number 7 (SS7) call control protocol, originally developed in the 1980's when the telephone network was migrating across to digital technology. It separated the paths taken by the voice and the call setup/teardown messages, so that a single signalling channel could control thousands of calls.
This protocol was used as the basis for GSM mobile phone networks and evolved into the GSM standard known as MAP and embedded in the 09.02 specification which stretched to many hundreds of pages. Its still very much alive and in use today, being maintained and updated by the 3GPP standards organisation. It also handles SMS test messages and controls access for prepaid users.
Data sessions are also setup and torn down using signalling messages. These are very similar for the GPRS data service used in both 2G and 3G networks.
With later releases of the GSM and UMTS standards, the path which the voice or user data traffic takes is more independent of the signalling control messages. Traditional voice switches known as MSC (Mobile Switching Centres) have been separated into MSC servers (which handle all the signalling and call control functions) and Media Gateways, which handle any adaption of the voice required at the boundary of the network. These do not need to be co-located as before.
There has been a remarkable change in traffic mix on mobile networks, where data has grown from 2% to over 50% of total traffic volume (although not revenue) in many 3G networks. The volume of data traffic has also grown astronomically, with predictions of anything from 10x to 100x over the next 5 years. This has justified a change in the underlying transmission networks, first to ATM, more recently to IP and which can be carried over high capacity transmission mediums such as Gigabit Ethernet and/or microwave links.
At this stage, the switching protocols, services and end-user experience have not radically changed over the last few years. Whilst costs have been reduced and data services have increased in speed and performance, the traditional services continue to be compatible with existing phones in almost every part of the world. The telecommunications network is probably the largest interconnected machine in the world, allowing over 3 Billion phone users to connect to any other individual user within a few seconds.
The reasoning for IMS
Once you have ubiquitous packet data communications, you can run many different services over it - voice, email, browsing, video etc. The internet today delivers a remarkable plethora of innovative and exciting services, and growing daily. However, these use what is termed a "best effort", shared approach. If your neighbours decide to watch a video at the same time as yourself, network capacity may be overloaded and performance for all is reduced. This is particularly concerning if you want to use an important service, such as a voice call, which would be seriously affected by delayed packets, resulting in unintelligible speech.
Fixed network operators today are tackling this problem using techniques such as DiffServ, where traffic is classified into (say) three or four priority levels. Business users and voice traffic is given higher priority, and thus receive better performance. However, there is still contention for the variable resource.
IMS is a complete core network specification which includes elements for call (or session) control, media gateways and service platforms. It adopts the SIP call control protocol. Once major benefit is that any session being setup can have a defined capacity allocated to it, which can change during the call. This allows voice, video or data sessions to be setup between end parties, something that the original circuit switched protocols weren't designed for.
Benefits include the ability to use higher data rates for voice calls - just as we have HDTV we could have HD voice, and to share the same common network and control elements for both voice and data. In today's cellular networks, these are separate voice and data switching systems in the core network which share the common radio access network.
Existing IMS networks
Although there are no full scale commercial IMS networks in service today to my knowledge, vendors have demonstrated products over recent years. The core signalling technology of SIP forms the basis for many VoIP (Voice over IP) networks, and handles billions of voice calls every day. It has been shown to interwork with existing phone systems, routing the calls over IP without the end user being aware. Many such systems form the basis for long distance or low cost international services. Indeed, ATT Mobile use SIP based VoIP to route all inbound calls to the appropriate mobile switching centre (MSC).
The transition to IMS
With over 3 Billion mobile phones using current phone call control messages, high compatibility so that most phones can technically work on most networks and a highly established commercial roaming system the current solution meets most user's needs.
IMS would require different phones (or at least a change to existing ones), offer a different set of supplementary services and use a different scheme to support roaming between network operators.
During a transition to IMS, operators would need to be able to cater for both old and new schemes. End users should not be expected to fully understand all the reasons why their service is changing, but would reasonably expect the same or better service experience and compatibility with other mobile users.
For example, telephone numbers in their current form would be supported, calls to voicemail handled as before and SMS text messages sent, stored and received as quickly as before. Additional services, and improvements in the end user experience would also be expected - whilst retaining compatibility or fallback when roaming to legacy networks.
Several vendors claim to solve this problem by providing core network elements that can interworks with both IMS and current mobile technologies. They translate the signalling messages to emulate current systems, talking the same protocols as other current systems such as MSCs. They can use existing telephone number schemes to address and route calls, in addition to other internet based addresses. Billing records can be output which can be translated into formats for roaming settlement purposes.
Whether the substantial costs and risks in making this transition are worthwhile remain to be seen. Whilst IMS was a popular topic at the 2007 3GSM World Congress, it seems to have lost its attraction. The primary benefits of moving to a common data transmission scheme are being achieved without changing the switching control protocols.
There needs to be further strong arguments made for the commercial investment to migrate across to IMS which do not appear to be made as yet. Perhaps we are likely to see the fixed network operators adopt IMS first, enabling them to guarantee quality of service for data services. This could then spread to mobile networks, initially for data services only, especially where an operator owns both fixed and mobile assets and shared a common core network between them.
IMS in femtocells
Several proposals exist for adopting IMS and its embedded SIP call control protocols with femtocells. These include:
- Complying with current standard mobile phones and the Iu interface, but internally converting all the messages into SIP based protocols. This would not give any visible benefits to the end user, add risk to the project and not provide compatibility between femtocells as required by network operators. This solution, advocated by Tatara systems, seems unlikely to be adopted.
- Complying with the Iu interface, but instead of connecting to a standard 3G core network, instead converting the Iu signalling into a common SIP core network. This solution, offered by vendors including Sonus networks, allows sharing of common services across wired and wireless phones. This would be particularly attractive in enterprise environments, where services aware of wired deskphones and mobile phones controlled by the same common network would provide intelligent call handling between both.
- Using a convergence server, which can operate with both IMS and current mobile protocols to support phones talking either language. Several vendors offer such solution, including Mavenir, Tatara, Alcatel-Lucent and Hitachi. These systems would require updated mobile phones to use the new IMS protocols directly, although they are also compatible with existing phones.
- Incorporating the IMS core network functionality into the femtocell itself, so it can operate autonomously almost as its own network. It can originate VoIP calls to other SIP capable networks and directly offload data traffic to the internet. Outbound calls do not need to be routed through the operator's core network, although incoming traffic using the mobile phone number would need to so do.
In some ways, this is similar to what can be achieved with WiFi and VoIP today. By signing up for a VoIP service, you can be allocated a fixed phone number enabling you to make or receive voice calls over your fixed broadband connection. Femtocells can translate your current cellular signalling into VoIP call requests and route them through a SIP server. When away from your femtocell, the call would be handled using the existing outdoor cellular system instead.
Once disadvantage is likely to be the poor call handover between femtocell and outdoor macrocell. A standard called Voice Call Continuity (VCC) has been developed to address this, but there are concerns about its performance. It requires setting up a new data session when leaving your femtocell zone which can take a few seconds, then handing the call over whilst still using VoIP rather than native voice service. This is likely to be more error prone and thus give a worse user experience.
Femtocells offer a way in which operators can experiment with this new system architecture in a controlled environment. They can choose which option to offer and whether to require new mobile phones at the same time or not. Where successful, it will lay down a migration path for adoption of the future 4G technology LTE, which is even more heavily data oriented than today's systems.
Some vendors are offering a SIP based architecture for femtocells, which is gaining traction in the Japanese market and of interest to CDMA operators. In the longer term, the 4G LTE system may migrate to this architecture. Depending on the specific architecture chosen, this may require new mobile phones, will have some compatibility issues, some aspects of poor performance and substantial migration costs. Whether the benefits of this approach justify this complexity and expense remain to be seen.