Sunday, 12 July 2015

S8HR: Standardization of New VoLTE Roaming Architecture

VoLTE is a very popular topic on this blog. A basic VoLTE document from Anritsu has over 40K views and my summary from last years LTE Voice summit has over 30K views. I assume this is not just due to the complexity of this feature.

When I attended the LTE Voice summit last year, of the many solutions being proposed for roaming, 'Roaming Architecture for Voice over LTE with Local Breakout (RAVEL)' was being touted as the preferred solution, even though many vendors had reservations.

Since then, GSMA has endorsed a new VoLTE roaming architecture, S8HR, as a candidate for VoLTE roaming. Unlike previous architectures, S8HR does not require the deployment of an IMS platform in VPLMN. This is advantageous because it shortens time-to-market and provides services universally without having to depend on the capability of VPLMN.



Telecom Italia has a nice quick summary, reproduced below:

S8HR simplicity, however, is not only its strength but also its weakness, as it is the source of some serious technical issues that will have to be solved. The analysis of these issues is on the Rel13 3GPP agenda for the next months, but may overflow to Rel14. Let’s see what these issues are, more in detail:


Regulatory requirements - S8HR roaming architecture needs to meet all the current regulatory requirements applicable to voice roaming, specifically:
  • Support of emergency calls - The issues in this context are several. For example, authenticated emergency calls rely on the existence if an IMS NNI between VPLMN and HPLMN (which S8HR does not provide); conversely, the unauthenticated emergency calls, although technically feasible in S8HR, are allowed only in some Countries subject to the local regulation of VPLMN. Also, for a non-UE-detectable IMS Emergency call, the P-CSCF in the HPLMN needs to be capable of deciding the subsequent action (e.g. translate the dialed number and progress the call or reject it with the indication to set up an emergency call instead), taking the VPLMN ID into account. A configuration of local emergency numbers per Mobile Country Code on P-CSCF may thus be needed.
  • ­Support of Lawful Interception (LI) & data retention for inbound roamers in VPLMN -  S8HR offers no solution to the case where interception is required in the VPLMN for inbound roamers. 3GPP is required to define a solution that fulfill such vital regulatory requirement, as done today in circuit switched networks. Of course VPLMN and HPLMN can agree in their bilateral roaming agreement to disable confidentiality protection to support inbound roamer LI but is this practice really viable from a regulatory point of view?
Voice call continuity – The issue is that when the inbound roamers lose the LTE coverage to enter into  a 2G/3G CS area, the Single Radio Voice Call Continuity (SRVCC) should be performed involving the HPLMN in a totally different way than current specification (i.e. without any IMS NNI being deployed).
Coexistence of LBO and S8HR roaming architectures will have to be studied since an operator may need to support both LBO and S8HR VoLTE roaming architecture options for roaming with different operators, on the basis of bilateral agreement and depending on the capability.
Other issues relate to the capability of the home based S-CSCF and TAS (Telephony Application Server) to be made aware about the VPLMN identity for charging purposes and to enable the TAS to subsequently perform communication barring supplementary services. Also, where the roaming user calls a geo-local number (e.g. short code, or premium numbers), the IMS entities in HPLMN must do number resolution to correctly route the call.
From preliminary discussions held at Working Group level in SA2 (architecture) and SA3 (security) in April, it was felt useful to create a new 3GPP Technical Report to perform comprehensive technical analysis on the subject. Thus it is expected that the discussions will continue in the next months until the end of 2015 and will overheat Release 13 agenda due to their commercial and “political” nature. Stay tuned to monitor the progress of the subject or contact the authors for further information!
NTT Docomo also did some trials back in February and got some brilliant results:

In the trials, DOCOMO and KT achieved the world's first high-definition voice and video call with full end-to-end quality of service. Also, DOCOMO and Verizon achieved the world's first transoceanic high-definition VoLTE roaming calls. DOCOMO has existing commercial 3G and 4G roaming relations with Verizon Wireless and KT.
The calls were made on an IP eXchange (IPX) and network equipment to replicate commercial networks. With only two months of preparation, which also proved the technology's feasibility of speedy commercialization, the quality of VoLTE roaming calls using S8HR architecture over both short and long distances was proven to be better than that of existing 3G voice roaming services.


In fact, NTT Docomo has already said based on the survery from GSMA's Network 2020 programme that 80% of the network operators want this to be supported by the standards and 46% of the operators already have a plan to support this.


The architecture has the following technical characteristics:
(1) Bearers for IMS services are established on the S8 reference point, just as LTE data roaming.
(2) All IMS nodes are located at Home Public Land Mobile Network (HPLMN), and all signaling and media traffic for the VoLTE roaming service go through HPLMN.
(3) IMS transactions are performed directly between the terminal and P-CSCF at HPLMN. Accordingly, Visited Public Land Mobile Network (VPLMN) and interconnect networks (IPX/GRX) are not service-aware at the IMS level. The services can only be differentiated by APN or QoS levels.

These three technical features make it possible to provide all IMS services by HPLMN only and to minimize functional addition to VPLMN. As a result, S8HR shortens the time-to-market for VoLTE roaming services.

Figure 2 shows the attach procedure for S8HR VoLTE roaming. From Steps 1 to 3, there is no significant difference from the LTE data roaming attach procedure. In Step 4, HSS sends an update location answer message to MME. In order for the MME to select the PGW in HPLMN (Step 5), the MME must set the information element VPLMN Dynamic Address “Allowed,” which is included in the subscribed data, to “Not Allowed.” In Step 6, the bearer for SIP signaling is created between SGW and PGW with QCI=5. MME sends an attach accept message to the terminal with an IMS Voice over PS Session Support Indication information element, which indicates that VoLTE is supported. The information element is set on the basis of the MME’s internal configuration specifying whether there is a VoLTE roaming agreement to use S8HR. If no agreement exists between two PLMNs, the information element will not be set.

The complete article from the NTT Docomo technical journal is embedded



5 comments:

Vijay Nagarajan (via GSMA VoLTE Linkedin group) said...

Let's see which way the deployment dice rolls; S8HR is a candidate but I have my doubts. Thanks for the article and summarising the pros as well cons around it.
I want to see in the real-world what if there would be call-quality and latency drop due to the fact that the codec/IMS processing may now happen miles or continents away.

Aleksey Makarov (via GSMA VoLTE Linkedin group) said...

This idea is shown passing in the IR65-v12. Perhaps it is good for short distances, but on intercontinental roaming will be big pads on the delay.

Ambrish Kumar (via GSMA VoLTE Linkedin group) said...

Also some country regulatory Mandates terminating media must go via T-H-PLMN. This why RABEL model came into existence. Non-OMR is required for such.

Unknown said...

...and when we come to monetize and reconcile the itsy bitsy packets the fun begins...

Nick Outram said...

The decision on what method to use for International IMS really should take a holistic view as "Unknown" is alluding to above the Commercial/Wholesale (IT) parts of s8HR are not so easy, it feels like s8HR solves some Network implementation issues only to create bigger ones in the commercial (/IT implementation) space.