Publié le 17 décembre 2019
In a recent discussion with friends about 4G and 5G, the topic of 5G network slicing came up. One of my friends correctly pointed out that although network slicing is not really a new concept, it will be key for 5G and beyond. Let’s examine what 5G network slicing means, why it is important, what kind of challenges it entails and how we can address them.
End-to-end network slicing is, arguably, the defining 5G feature. The concept of slicing–partially implemented in mobile networks up to now, as domain specific (e.g., core) slicing–is not new. 5G network slicing splits resources into logical or virtual networks (“slices”) to address use cases with distinct characteristics and service level agreement (SLA) requirements.
A 5G slice that supports a critical IoT use case such as robot automation would differ in terms of throughput, latency and reliability requirements from a 5G slice for a typical voice phone call—assuming that people will still call each other in the future. Of course, individual slices must be considered together and in the light of all available network resources to ensure that each slice provides users/devices with the relevant service quality and end-to-end network performance. In other words, the 5G network may sometimes have to “politely” turn down requests for another slice.
As, arguably, the defining feature and “secret sauce” of 5G, end-to-end network slicing is fundamental for the 5G success. This explains why so many conference presentations, webinars, white papers, articles and blogs (including this one) focus on slicing in 5G. Without network slicing, the 5G promise of flexibility and optimal resource efficiency–and therefore maximum return on investment (ROI)–cannot materialize.
Similarly, the implementation of new business models built upon mobile network-as-a-service (NaaS) would be incomplete without slicing. This may also affect the 5G ability to support mission-critical and network-performance-sensitive use cases for new verticals, including industry automation (Industry 4.0) and health applications. Even a partial consideration of slicing in 5G is likely to negatively impact the adoption and revenues from new, enterprise-focused services. So, yes, one could say that 5G network slicing is the greatest thing since sliced bread.
Albeit promising, 5G network slicing comes with significant challenges. For slices to support use case specific requirements, the underlying virtual/cloud and physical infrastructure is crucial too. For example, fiber network issues or suboptimal radio coverage would directly affect a 5G slice.
The need for accurate end-to-end view of network performance, from the user/device to the data server, is also challenging, particularly as real-time visibility is mandated for many 5G applications. Of course, such visibility should also focus on what users/devices “perceive” as network performance, which remains a substantial challenge in mobile networks. Finally, the dynamic orchestration (coordination) of 5G slices so that these are optimally created and maintained (based on network resource availability and use case requirements, potentially by deprioritizing less critical applications/services at certain times) should not be underestimated. Even with slicing in 5G, there is no gain without pain.
5G network slicing necessitates effective preparation and monitoring. Before slicing is commercially available, every part of the 5G architecture should be slicing-ready. Specific domains (access, core, transport) as well as the underlying virtual/cloud or physical (e.g., fiber) infrastructure must be tested and, if needed, upgraded. The network/service topology should also be accurately known. Such knowledge is vital in operations too, as the 5G network is dynamically partitioned into slices.
Real-time active and passive monitoring will also be pivotal for the–optimal, if not mandatory–user/device-centric, analytics-driven and automated assurance approach to network slicing. And we should not forget the essential role of machine/deep leaning and artificial intelligence (AI), for example in predicting and preempting issues that could affect users/devices or in intelligently automating slicing-centric processes as part of the overall 5G network orchestration. In summary, to address 5G network slicing challenges, both the trees and the forest matter.
5G network slicing requires optimal domain-specific and cross-domain, end-to-end performance on top of the underlying hybrid infrastructure
(CU/DU: centralized/distributed unit, IMS: IP multimedia subsystem, IoT: internet of things, MEC: multi-access edge computing, UHD: ultra high definition, V2X: vehicle to anything)
This blog is only a high-level introduction to the complex topic of 5G network slicing. For more information about slicing in 5G, you can watch our relevant webinar. For more details about the 5G architecture and pertinent concepts such as network function virtualization (NFV) and software-defined networking (SDN) or about 5G deployment and operational challenges, please consult EXFO’s website or contact EXFO.
You may also be interested in our new podcast, “Know your network,” featuring thought-provoking conversations around key management challenges facing today’s carriers.