GMI 2006: Lessons Learned

By James McEachern

Introduction

GMI 2006 also validated the MultiService Forum (News - Alert) Architecture Release 3 G lobal MSF Interoperability 2006
(GMI 2006) represented the first
international, multivendor validation of IMS ( IP Multimedia Subsystem (News - Alert) ), the underlying framework that will enable true service convergence. GMI 2006 validated the separation of services from access technology, which allows subscribers and devices to access the network from anywhere, eliminating the boundaries that separate fixed, mobile, and IP networks. This is the essence of Fixed-Mobile Convergence (News - Alert) . as a full peer network to a “pure” IMS implementation.MSF R3 architecture includes third-party applications, service brokering, enforceable QoS, and interworking between the PSTN and IMS. The MSF architecture also includes support for advanced services, such as priority calling as specified for Emergency Telecommunications Service (ETS). GMI 2006 demonstrated a practical, realworld deployment scenario encompassing the PSTN, deployed first generation VoIP networks, and emerging IMS networks.

GMI 2006: Why Now, Why the MSF?

Did the world really need another interop event? The answer, of course, depends on exactly what type of interop event we are talking about. An event focused on a single interface might have been useful, but it would not have significantly enhanced the industry’s understanding of the maturity of IMS. On the other hand, an event that assembled the key pieces of a real world carrier network would be a different matter. If this event combined IMS with deployed VoIP systems and third party application servers, on a globally networked test bed, it would begin to resemble what carriers will actually see in their networks as they deploy IMS. An interop event like this would bring real value to carriers who are in the process of making decisions about the deployability of IMS. That is why this is exactly what GMI set out to do.When the plans were being drawn up for GMI 2006, marketing preferred to see all the tests succeed. The engineers wanted to see how many tests they could make fail. The engineers won, and were given free rein to identify as many problems as possible.

Why was this important interop event hosted by the MSF? Because the MSF is one of the very few organizations with the technical expertise, the end-to-end architecture, and the practical experience, to successfully execute an interop event of this scale.

GMI 2006 was conducted over a 12-day period from October 16 through October 27, at major carrier and independent labs on three continents, networked together for this event. Five of the world’s top carriers—BT, KT,NTT, Verizon (News - Alert) and Vodafone—along with world-class testing and research facilities at the University of New Hampshire Interoperability Lab (IOL) and ETRI, provided the host sites for this interop event, sponsored by Nortel (News - Alert) . A total of 26 vendors brought 200 pieces of equipment to GMI for testing. The testing was conducted by some 200 engineers at the GMI host sites. For every engineer at one of the five host sites, additional engineers supported them from their company labs. In total, well over 500 engineers were involved in GMI 2006, working 14 or more hours a day, testing MSF Release 3 Implementation Agreements covering a wide range of topics including roaming, QoS, and network interconnect.

The breadth of GMI, particularly the end-to-end global validation, was a critical aspect of this event. By evaluating the end-to-end operation of IMS networks and services, GMI 2006 demonstrated that IMS is ready for real-world networks now.“Interoperability is the key to the transition to IMS,” explains Roger Ward, Office of the CTO, BT Group (News - Alert) , and President of the MSF.“In practice, carriers have networks at various stages of IMS and NGN implementation.We see networks with a mix of legacy infrastructure and pure IMS gear, and a broad array of multivendor equipment. The MSF and its GMI validations are concerned with practical, real-world considerations and explicitly address the heterogeneous environment that exists in carrier networks today.”

GMI 2006 was an unqualified success, validating IMS and the overall MSF Release 3 Architecture. Specific issues were uncovered, and will be used to help improve standards. The insight from GMI 2006 highlighted what works well and revealed where more still needs to be done. Key conclusions from GMI 2006 will be forwarded to the Standards Development Organizations (SDOs) responsible for the standards at the core of IMS, allowing these standards to be refined where appropriate.

What did GMI test?

GMI 2006 was conducted in a series of eight scenarios that began with a basic configuration, and then progressively added functionality. Building up the complexity in this fashion ensured that GMI would provide the maximum possible insight into what worked, what didn’t work, and why. The initial scenarios established the underlying connectivity across the network, and verified that this network could be automatically established, tested, and reconfigured as required. The next scenarios tested basic services within a single IMS domain, and then added additional functionality until the final scenario, which included roaming across multiple IMS domains, providing basic and value added services, with guaranteed QoS. The final scenario is illustrated in Figure 1.

GMI 2006 Test Results As GMI testing progressed, signalling flows were collected for each test, allowing full diagnosis to be performed both during and after GMI. The full results have been published in a GMI white paper, which can be accessed on the MSF website at www.msforum.org/interoperability/wp2006.shtml. The key results and conclusions are summarized here, but those interested in more detail are encouraged to download the GMI White Paper (News - Alert) from the MSF.

The test results from GMI 2006 showed a level of maturity in IMS standards that many had not expected. Core IMS products worked virtually out of the box, in various multi-vendor configurations.Most of the problems encountered were in the components around the core IMS. For example, in some cases, Application Server’s SIP stacks were not in full compliance with the SIP standards. These issues were identified quickly, and the problems fixed. In most cases, these tests were re-run and completed successfully.

There were also issues that could not be resolved during GMI, or that pointed to underlying problems that needed to be addressed in standards bodies. These issues fell into three broad categories.

Functional to Physical Architecture Mapping: IMS defines a functional architecture, with vendors having some freedom in how these functions are combined into actual implementations. The MSF aims to produce a more physical architecture, with recommendations for preferred mappings from functional to physical, supported by the appropriate Implementation Agreements. In one case the recommended physical implementation combining the policy function with the P-CSCF was not implemented by any vendors, and the MSF architecture will be updated to reflect this. GMI also highlighted the need for decomposed Session Border Gateways (SBGs — sometimes also referred to as Session Border Controllers) with separate control and data functions connected through standard interfaces. This was supported by some, but not all, SBG vendors.

Standards: GMI provided valuable insight into a number of potential standards issues. The data is being analyzed, and contributions to the appropriate standards bodies are being prepared. In some cases, the data suggests possible weakness in the standards. For example, in some tests the SBGs were modifying SIP route headers at the NNI in ways that worked for network-to-network traffic, but that led to call failure during roaming scenarios. Further investigation is required to determine whether this was caused by a simple implementation error, or a problem with the standards. Another example was RFC 3312 (Preconditions), which is an important standard for end-to-end QoS. GMI found that all the implementations that were tested appeared to deviate from the standard, leading to inconsistent behavior. This may indicate a fundamental problem with the standard, and the data from GMI will be used to provide feedback to standards. Finally, there were cases where standards were simply not being implemented. The MSF defined a mechanism to allow optimal routing of traffic during roaming to support improved QoS, but no implementations were available during GMI. Further investigation is required.

Gaps / Overlaps: GMI also identified areas where the problem was not a lack of standards, but rather too many standards, and too many options within those standards. One such area was authentication of users in IMS. Once the various configuration options were sorted, authentication worked as intended, but the necessary configuration was complex and time-consuming. Clearly it is not realistic to expect the average user to deal with this level of complexity. This seems to be a problem ideally suited to an Implementation Agreement that would narrow the range of options to improve interoperability. The MSF is evaluating the possibility of developing a new IA addressing authentication/authorization profiles.

Over the coming months liaisons will be sent to IETF, 3GPP and TISPAN detailing the problems identified during GMI 2006, and providing data to support our conclusions.We expect that the ensuing dialog will identify solutions which the industry can collectively pursue. The result will be a further strengthening of the IMS standards, and the products based on these standards.

Finally, GMI highlighted some areas where it may be appropriate to consider going beyond basic interoperability tests, and moving into the domain of formal certification testing. Two initial candidates being considered for certification testing are the NNI interface between networks, and the UNI interface between User Equipment and the IMS network. The MSF is currently gauging the level of interest and investigating the technical feasibility of certification in these areas. This represents an exciting new direction that may complement the work being done in GMI.

Conclusions

The single most important lesson learned from GMI 2006 was that IMS is more mature and demonstrating greater multi-vendor interoperability, than many in the industry expected. Some naysayers have suggested that IMS is too large and too complex to ever achieve wide-scale commercial deployment. GMI 2006 has clearly shown this is not the case, and that the reality of deployable, carrier-grade IMS is much closer than the skeptics claim.With little more than straightforward configuration, much of the equipment was up and running, in multivendor deployment scenarios, within a matter of hours. This was especially true of the core IMS, where multivendor interoperability was the norm. Even complex advanced services functionality, such as interworking of priority calling for Emergency Telecommunications Service (ETS) between PSTN and IMS domains, was successfully demonstrated. IMS isn’t quite at the “plug and play” stage yet, but it is certainly on a par with much of the existing telecom and internet infrastructure.

Of course, this does not mean that everything worked exactly as specified in all configurations. There were test cases that failed. In these instances, detailed call traces were collected and analyzed. Many of the problems were eventually diagnosed as simple implementation errors — in other words, software bugs. GMI provided these vendors with valuable product testing. Other cases pointed to areas in the standards that may need to be revisited. But even in these cases, the issues were generally not problems with the core IMS specifications. Rather they pointed to the infrastructure around IMS, such as value added application servers, and the mapping of the IMS functional architecture to the MSF physical architecture. The problems encountered were consistent with ongoing optimization of a stable standard, and were inherently addressable. The feedback from GMI is exactly the input needed to address these issues, and concrete action plans have been drawn up in the MSF technical committees to do this. Feedback is being provided to the Standards Development Organizations (SDO) responsible for the relevant standards, and will contribute to further improvements in IMS interoperability.

GMI 2006 has shown that IMS is mature enough for practical multivendor deployment to begin. Interoperability events like GMI will continue to improve IMS to achieve the full potential of converged services.

About the MSF

The MSF is a global association of service providers, system suppliers and test equipment vendors committed to developing and promoting open-architecture, multiservice Next Generation Networks. Founded in 1998, the MSF is an open-membership organization whose members are drawn from the world’s leading telecommunications companies. The MSF’s activities include developing Implementation Agreements, promoting worldwide compatibility and interoperability of network elements, and encouraging input to appropriate national and international standards bodies.

MSF is a well-established forum with a balanced mix of carriers and vendors that integrates specific work from multiple standards into a holistic network and services architecture. The MSF architecture and solution framework combine legacy and next-generation services in a single unified network. Further, since all MSF participants implement the same baseline features and functions, members can eliminate the guesswork that technology development typically involves.

James McEachern is Vice President MSF & Nortel, Carrier VoIP Standards Strategy. Visit Nortel Networks online at www.nortel.com.