Elementary, my dear Watson
Early VoLTE rollouts presented a number of challenges for service providers, including VoLTE call drops due to timeout. This particular issue appeared to be IMS related, but the Core/IMS team could not identify its root cause. IMS had just been deployed, so the inability to pinpoint exactly where the issue lay was attributed to lack of knowledge.
The mystery was solved after considering data from the radio access network (RAN). It then became evident that the VoLTE call timeout was a RAN issue, caused by noncontiguous radio coverage when users moved between cells. To resolve the matter, the RAN and Core/IMS teams simply had to communicate and correlate data across network domains. Elementary, one would say.
Well, this is easier said than done. As discussed in “Start with the end in mind”, service providers need to better manage how mobile users experience the end-to-end network performance. This need has been accentuated by the evolution of mobile networks from circuit-switched, voice-centric to packet-switched, data-centric.
But how can you manage network experience unless you are able to measure it accurately? The inadequacy of ‘traditional’ metrics or key performance indicators (KPIs) has been discussed for years. Although cell-averaged (performance management) data and drive/walk test measurements (for specific areas, times and days) are valuable, they have inherent limitations.
Relying on customer complaints can also be misleading. After launching VoLTE, some service providers received few complaints about their updated ‘green button’ support for mobile voice. But this was due to the slow VoLTE uptake and – in some cases – user dissatisfaction leading customers to revert to legacy or over-the-top voice offerings.
This is why many service providers have added other data sources to their armory. For example, call traces can drive subscriber/device focused analysis, including experience geo-location on a granular (street and building) level. The requirement for end-to-end network visibility has also led to the targeted adoption of probes, whether passive or – in the absence of mobile traffic – active. Additionally, customer and business information has started to permeate network planning/optimization.
Understanding what each KPI implies would also be key. What is the experience of mobile users in an area with 100% cellular coverage by service provider X? Could a competitor do better despite covering only 95% of the same area? The area characteristics, base station and antenna placement, network capacity, number of users and user profile are crucial factors in determining network experience. So, 100% coverage is an important but not sufficient metric. And there are cases where the experience of users in perfect-coverage networks cannot match what seemingly worse networks provide.
In fact, the VoLTE case described at the start is relatively straightforward. To identify the root cause of the call timeout and similar network experience issues, access to all relevant data across domains is mandatory (an open communication channel between network teams would help too). In addition, the significance of impartiality, beyond vested interests and domain or network equipment specific biases, should not be underestimated.
Of course, there are many other and even more intricate network experience challenges to address. Such challenges call for Sherlock Holmes like powers of observation, intuition and deduction to ensure effective and efficient analysis. But how can engineering/operational teams cope with complexity and the tsunami of structured/unstructured data? Well, the essential and costly asset of network expertise does not have to be limited to humans. Network analytics can be of great value too.
To make sense of data in a timely manner, expert software can assist with fast cross-domain correlation and analysis automation. Such an intelligence augmentation (or amplification) approach becomes a must in live network operation. Actually, connecting the network experience dots should not be reduced to a reactive, offline or real-time analysis. Moving to a proactive and eventually predictive mode of managing how users experience network performance will not be optional for much longer.
And what about artificial intelligence, which has been dominating news headlines? Complementary, my dear reader (and a great topic for another blog)…
Mobile World Congress Barcelona 2017 is approaching fact. Check the main Amdocs Network Solutions highlights at this year’s event.
Author: Dr. Konstantinos Stavropoulos is responsible for product marketing in Amdocs Network Solutions, leading the definition and marketing of productized RAN and cross-domain network offerings. Konstantinos has been focusing on mobile networks for more than 20 years through academic and professional roles. His experience covers diverse areas, from antenna array systems research and mobile network planning/optimization consultancy to the development, management and marketing of innovative software solutions.