Dynamic network slicing has been a 5G service capability from the earliest days of 5G development. The slicing concept is relatively straightforward. Instead of a traditional ‘one-size-fits-all’ approach towards resource allocation, network resources are dynamically allocated using virtual networks optimized for individual services. For example, a network slice (i.e., virtual network) to support IoT might have sparse 5G core resources and no handover functionality but massive connectivity demands. In contrast, a mobile broadband slice is likely to require extensive 5G core resources, fully-featured handover capabilities, and support high data rates.
At first glance, dynamic network slicing resembles the quality-of-service (QoS) classes introduced in 3G with limited success. Arguably the Achilles heel for 3G-QoS was the additional infrastructure resources needed to ensure adequate QoS service levels. Slicing aims to avoid this trap using the native network virtualization capabilities inherent to 5G. Slicing also capitalizes on the diversity of digital service demands as enterprise and consumer use-cases proliferate.
Dynamic network slicing should be coveted in a world besieged with network commoditization. But it isn’t. In this report, we summarize some of the reasons why. We investigate what needs to change, the long-term prospects of slicing, and the implications for other digital ecosystem stakeholders, such as enterprise end-users, device manufacturers, application developers, and cloud and edge computing infrastructure providers.
When 5G was first deployed in South Korea and the United States in April 2019, it came nearly a year ahead of schedule, buoyed by competitive market forces and geopolitical wrangling. Compromises with technology standards and network upgrades were necessary to deliver 5G on this accelerated schedule. These compromises emphasized 5G radio technology developments to enable 5G enhanced mobile broadband (eMBB) capabilities using 4G core network equipment in so-called non-standalone (NSA) architectures. Most 5G deployments are currently NSA-based. 5G core network upgrades with so-called standalone (SA) architectures are necessary for network slicing.
SA network upgrades are driven primarily by market demands and, in some cases, unique technical considerations. For example, T-Mobile in the USA has aggressively upgraded its network with a 5G-SA core because of its unique 5G radio spectrum at 600MHz, which provides umbrella coverage as opposed to supplementary capacity in the case of a typical NSA configuration.
When network operators deploy 5G-SA, they can include dynamic network slicing capabilities, but that isn’t the end of the story.
Slicing aims to capitalize on virtualization and cloud operations to allocate network resources according to service demands dynamically. However, network operators must fundamentally transform their network architectures and operations before being virtualized and cloudified. Network architectures require software-defined and virtualized compute functionality (with ‘five-nines’ performance) strategically scaled and distributed across their network footprints. Operational organizations must be transformed, and staff retrained to support new processes and procedures for virtual infrastructure. In addition, new operational support systems with sophisticated automation capabilities are needed to manage and orchestrate network virtualization.
Early pioneers, including AT&T, NTT, Orange, and Verizon, have spent many years transforming their network operations to support network virtualization. In 2016, AT&T formalized its efforts with the ECOMP initiative , which subsequently became the Linux Foundation’s ONAP project underpinings. A variety of software and technology vendors, including Amdocs, Ericsson, Fujitsu, Huawei, NEC/Netcracker, Nokia, Samsung, and ZTE, contribute to the ONAP project and have sophisticated management and orchestration systems for virtualized networks. Nokia and Ericsson are the most vocal in promoting 5G network slicing. Both vendors have robust network equipment and operational software platforms well-positioned to capitalize on network slicing opportunities. Huawei is also a strong advocate for network slicing, but its global 5G aspirations are hindered by constraints imposed by the United States and other like-minded nation-states.
Once Tier 1 operators like AT&T stabilized their virtualized core network operations, they benefited from tremendous operational cost savings. For example, AT&T reported 17 quarters of incremental cost savings from virtualization, ranging between 7 and 8 percent year on year. These reported cost savings, coupled with improved virtualization technology and software and operational blueprints, are accelerating market demand for network virtualization. As a result, we expect network virtualization equipment and software revenues to increase by 25-30 percent over the next five years.
Virtualization is a journey for incumbent network operators, often starting with non-critical core network platforms and migrating through other core and transport network functions. As a result, most 5G radio access networks (RAN) use traditional platform-based architectures that are not well suited for dynamic network slicing.
As operational models for virtualized networks mature, RAN networks will inevitably be virtualized and cloudified. New entrant operators like Rakuten and Dish Networks are disrupting the status quo by deploying virtualized RAN platforms from the outset. However, we believe that it will take 24 to 36 months before virtualized and cloudified RAN platforms become more widely accepted amongst Tier 1 operators and improve the value proposition for network slicing.
While mass-market 5G adoption is the domain of public network operators, 5G is also heralding a groundswell in private networks. Private networks have been hindered by insufficient radio spectrum and sub-scale device and network ecosystems in the past. However, government policy makers in many markets have allocated mobile spectrum resources for private networks in recent years. For example, in the United States, CBRS spectrum has been allocated for shared public/private use. In addition, both C-Band and millimeter wave spectrum have been allocated specifically for industrial applications in other markets.
Since Private 5G enables dedicated networks for targeted services, they will displace network slicing opportunities for public networks. However, we do not believe Private 5G is necessarily the death knell for network slicing. Since most private 5G networks will be deployed greenfield, with less scale and targeted use-cases, they will be well suited for end-to-end virtualization from the outset. Network slicing will be easier to implement in cases where tiered performance capabilities are needed to support different digital services. For example, network slicing is proposed for industrial networks where best-effort services operate alongside time-sensitive (TSN) functionality.
It is not enough to focus on Tier 1 operators when evaluating the prospects for dynamic network slicing. If industry pundits play their cards right, we believe that private 5G can be a crucial driving force for dynamic network slicing in both private and public networks. Private 5G lowers market barriers and increases the potential influence of other ecosystem stakeholders with vastly different drivers and challenges with network slicing.
Optimal resource scheduling will be an ongoing challenge for dynamic network slicing and depend on sophisticated and largely proprietary multi-objective optimization algorithms and complex management and orchestration platforms. End-to-end coordination amongst network functions will be challenging and constrained by operational silos and a lack of standardization for the foreseeable future. As a result, initial network slicing solutions will likely have course resource allocation schemes with modest dynamics. For example, some public operators have considered initially using slicing for internal network optimization efforts before launching customer-facing solutions.
A measured approach to dynamic network slicing is essential for both private and public 5G networks. This is particularly the case as 5G network traffic increases with diverse service demands that have ‘bursty’ traffic profiles. Poor system performance is likely when network slices are too granular for the dynamic resource algorithms to cope.
Although dynamic network slicing has been touted for many years, there is still a lot that needs to be done before its mainstream adoption. Most notably, 5G networks must be upgraded with 5G-SA core network capabilities, and radio transport and core network functions must be virtualized and cloudified. While this is likely to take 24 to 36 months for virtualization to become commonplace in public networks, it is already the preferred core network option and is fast becoming the preferred radio network option for Private 5G.
While Private 5G will cannibalize slicing opportunities in public networks, we believe that this will be offset by the role Private 5G plays in enabling broader ecosystem stakeholder engagement beyond mobile operators. However, since slicing involves complex resource optimization and orchestration, slicing solutions will likely have humble beginnings, with compatible resource granularity and dynamic allocation schemes.