The 3GPP standard body develops and releases all standards pertaining to the GSM/UMTS and LTE mobile network. Both of the 3GPP SA1 committee (which produces requirements specifications) and 3GPP SA2 (which looks after the system architecture aspects) have been dealing with this hot topic.
The Femto Forum (not a standards body itself) has hosted some very lively debate on the subject, with a view to achieving industry consensus so that time spent at standards meetings can focus on developing and approving the standards. This will reduce the time taken to reach an approved standard, and limit the risk of developing or deploying pre-standardised solutions.
The four main alternatives for SIP/IMS femtocell architectures
There have been several alternative proposals for the SIP/IMS architecture for UMTS femtocells, with different vendors proposing SIP user agent clients be hosted in almost every part of the transmission chain. Some 6 variations of SIP/IMS femtocell architectures have been submitted, some have been withdrawn and others added over time. They can be categorised based on where the SIP client is located:
- inside the femtocell access point (Home Node B)
- in the femto gateway
- outside the Home Node B subsystem (in a convergence server similar to the 3GPP2 CDMA femtocell architecture)
- nowhere (and instead use the existing ICS [IMS Centralised Services] standard to convert from circuit switch to IMS)
The arguments in favour of each option relate to the goal being targeted.
Incorporating a SIP client within the femtocell is better for local breakout
Where the purpose is to offload traffic at the femtocell access point (known as Local Breakout), then a more radical approach of incorporating the SIP client within the femtocell is appropriate. This is seen as highly conducive in the enterprise/business environment, where calls can be handed off to local IP-PBX voice switches and local data servers.
Incorporating a SIP client within the femtocell gateway is better for 2G/3G service compatibility
Where compatibility with existing core network services and features is paramount, then retaining the Iu-h interface and converting to SIP within the femtocell gateway is arguably more appropriate. This would not require major redesign/development costs within the femtocells themselves.
What we are seeing at the Femto Forum meetings, where these debates are being held, is a growing majority view:
- In the home: Use the existing Iu-h standard, with a SIP agent embedded in the femtocell gateway for future compatibility with IMS core networks.
- In the enterprise: Further need to identify the requirements. This is likely to break-out calls to the IP-PBX in a corporate environment: External calls would use Iu-h whilst local calls would breakout using SIP.
Using a Convergence Server to interwork between SIP/IMS and 2G/3G core networks
This is the same approach used within the 3GPP2 standards for CDMA femtocells, as we reported recently. This architecture applies the same principles, naming the box HNB CAS (Home Node B Convergence Application Server). The Iu-h interface and femto gateway are preserved for packet switch traffic, but circuit switch traffic is carried directly to the IMS core. New software and protocols within the femtocell work together with the HNB-CAS to provide this offload of voice traaffic from the legacy core network. The HNB-CAS looks to the 2G/3G network like another MSC/VLR and to the SIP/IMS network like a SIP Application Server.
Whilst this would provide some commonality with the CDMA femtocell architecture and move the industry further down the SIP path, it would be a radical change relative to the existing Iu-h solution (particularly for an operator otherwise undecided about IMS). Some operators are said to be very concerned about hosting another MSC/VLR box (from another vendor) in their network, probably due to implications around costs and timescales for testing/maintenance/configuration and compatibility.
Other standards are already prepared for a local breakout approach
As discussed in an earlier article, the 3G CDMA architecture defined by 3GPP2 already has a SIP client embedded within the femtocell. This architecture is ready for a local breakout option.
For LTE, the new 4G standard common to both CDMA and UMTS , femtocells are simply handling broadband IP. The various architectures being discussed at the macro level (i.e. not femtocell specific) for voice over LTE don’t directly affect the femtocell, which would continue to pass IP through to the handsets. Where there are dual mode 3G/4G femtocells in the future, there may be some commonality/benefit from combining the SIP user agent in the femtocell access point.
3GPP femtocell standards release timeline
3GPP Release 9, due out around the end of 2009, will include the final security/authentication aspects for UMTS femtocells which had been left out of the original Release 8 specifications. The same authentication procedures and scheme have been used for both UMTS and CDMA femtocells – adopting the same security model, certificates, IPsec and IKEv2 in both standards.
Work is also intense in the TR-069 remote device management protocol. This was recently extended with the TR-196 object model, and is further being refined with an amendment due by mid 2010.
3GPP Release 10 is due out in mid 2010 and will be a short release (i.e. limited content/short timeframe). It is very unlikely to include the full standard for enterprise femtocells although it is expected to have the enterprise femtocell requirements agreed by early next year. The content and schedule for Release 10 are not yet agreed.
We’ll probably have to wait until 3GPP Release 11 for a full standard incorporating IMS based enterprise femtocells, expected sometime in 2011.
A growing consensus?
What has been productive in recent weeks is the apparent growing majority support for the architecture regarding where the SIP client should be located in the IMS architecture for UMTS femtocells (femtocell gateway for domestic femtocells, femtocell access point for enterprise). Whilst many forum members support this view, there are some (including SoftBank and NEC of Japan as well as Tatara Systems) who strongly disagree – they would like to see the SIP client located in the femtocell.
The next decision point is the 3GPP SA2 committee meeting in Mexico in November where this architecture committee will debate these options and may adopt a working assumption.
The initial standard for 3G UMTS femtocells, which was released earlier this year, is being further refined and clarified around aspects for security and remote management. The focus of standards work is now shifting to the alternative SIP/IMS based architecture, with lively debate between the many different alternatives proposed.
There remains two broad camps for SIP femtocell architecture for GSM/UMTS based around where the SIP client should be based - in the femtocell or in the network.
Assuming a consensus around the SIP femtocell architecture becomes accepted, femtocell vendors are able to proceed with their investments knowing that the architecture is stable and that their efforts won’t be wasted. Operators can also confidently proceed with market trials and launch, in the knowledge that any pre-standards equipment shipped can be easily upgraded to the common standard when finalized.
Debate still continues however, with further requirements gathering for enterprise applications (which would justify SIP clients in the femtocell) likely to be the next practical step.
My thanks to Doug Knisely, VP Technology for Airvana, Femto Forum board member and Phil Brown, Standards Architect for Tatara who are both active participants in the standards bodies for providing these insights.