Your browser is not supported. Please update it.

24 November 2020

A1 Austria works towards automated 5G network slicing with Nokia

Network slicing has so far promised much but delivered relatively little in the way of benefits to operators in revenue or cost savings. That is not surprising for a technology in its infancy, although overhype has raised expectations prematurely at risk of disillusionment setting in. Nokia, despite or perhaps because of being slower off the blocks than certainly Huawei on this front, has been active in promoting network slicing this year as a crucial differentiator for 5G.

Network slicing in fact evolved under 4G through various options for partitioning networks into virtualized segments, underpinned by software-defined network (SDN) and network functions Vvrtualization (NFV) technologies that were becoming available. But 5G will enable greater flexibility and scope for network slicing with ability to differentiate far more between use cases in latency, capacity, speed and massive device scalability. Network slicing also becomes more complex to orchestrate and manage under 5G, which is why automation will be that much more critical.

That is why automation featured prominently in A1 Austria’s announcement last week of use of network slicing in its deployment for the country’s rail operator, ÖBB Infrastruktur AG, for onboard signaling and control. Currently this is running over A1’s LTE  network for transmission of location data and signals to traction units, enabling location of trains to be determined more precisely.

ÖBB said this spans the whole infrastructure from RAN through backhaul to core, utilizing automation features for 4G and 5G network slicing announced by Nokia in February 2020. This was delivered via a software upgrade to the existing LTE infrastructure, ready to progress to 5G Non-Standalone (NSA) and then on to 5G Standalone (SA) operation.

Nokia somewhat disingenuously claimed a world-first for that new software-based network management, control and orchestration, on the basis that the degree of automation was higher than had gone before. Such claims should be reserved for clear cut advances rather than the normal game of leapfrog, but at least Nokia is now better placed in the slicing arena.

Huawei can claim credit as the early runner after demonstrating the world’s first 5G network slicing application for smart grids in Beijing in 2017, featuring automated customization of slices to service requirements. It was also an early showing of the ultra-low latency use case, here for restoring power within 300ms to an area where it had been temporarily lost because of line damage, or some other reason.

Since then Huawei has sought to capitalize on that early lead through the 5G Slicing Association (5GSA), whose third annual summit held online from Beijing in July 2020 it co-hosted with China Mobile. Huawei played a key role in the largest commercial 5G SA network to date alongside all three major Chinese carriers, China Mobile, China Telecom, and China Unicom, which is now being employed by manufacturers and also utilities for smart grid and other applications.

There were several illuminating aspects of that summit, a major one being the clear identification of five categories of network slice, oriented almost on a horizontal basis matched to service requirements, rather than vertical industries. Two categories applied to consumer services and two to enterprises over public networks in each case, while the fifth deals with E2E (enterprise-to-enterprise) cases.

  • The category referred to as Level (L) 0 comprises basic B2C services by default, such as Internet access and best-effort video.
  • Category L1 is then preferential or ‘VIP’ public services, such as accelerated gaming and high quality video.
  • Category L2 is common industrial or business services such as extended reality (ER) and education, requiring more than basic best effort but no absolute guarantee of QoS.
  • L3 is then preferential or ‘VIP’ industrial services offering QoS guarantees for business parks, hospitals and manufacturing plants, but perhaps not ultra-low latency.
  • L4 is the E2E category called ‘industrial network special’, offering ultra-low latency and high levels of availability and support suitable public security and electric power provision. L4 is the only layer that offers total independence with full logical or physical isolation of resources.

The 5GSA is essentially a Chinese group and reminds us that there are still different architectural approaches on the table for network slicing. But the underlying principles of logical separation through virtualization enabled by SDN to separate the user plane from the control plane, along with NFV to enable functional isolation from underlying commodity hardware, is well-established.

Ericsson has made some valuable contributions in network slice operation, particularly on the IoT front. Ericsson’s IoT Core enables allows general mobile broadband traffic to be provisioned and billed as separate commercial services, with the IoT deployed as a network slice comprising the relevant virtual network functions (VNFs) such as the Serving Gateway, Policy Control and Charging Rules Function.

Ericsson says its latest packet core products have been designed from the ground up for network slice deployments with scalability in mind to give operators an affordable migration or upgrade path. The different network functions, some of which will often be common to multiple slices and others not, can be integrated or decoupled, running on shared or separate pools of hardware.

This is based on the concept of the Dedicated Core Network (DCN) or DECOR, introduced by the 3GPP in Release 13 to separate single core networks for different services dedicated to specific groups of subscribers or devices with common needs. This then evolved into enhanced DECOR (eDECOR) in Release 14, which reduced signaling and therefore latency to suit more delay sensitive slices, but at the cost of requiring upgrades to user equipment (UE).

Ericsson also supports differentiated QoS levels for critical applications and network slices at the RAN level with various options, such as specifying a minimum bandwidth per network slice, which can be different in each case. Other features include flexible counters that monitor QoS levels separately for each network slice to ensure adherence to service level agreements (SLAs) as far as possible.

There are also features for coping with congestion at the RAN level, with support for resource pre-emption where air interface resources are scheduled in advance for high priority services. Reservation can be dynamic where resources reserved for a given slice can be reused by other slices to varying extents. For highest priority slices static reservation may be employed where resources cannot be allocated to any other slice at any time under any circumstances. This comes closest to a traditional connection-oriented service over a dedicated end-to-end connection.

Network slicing will evolve to become more effective at optimizing use of spectrum while meeting radically different QoS requirements, but ultimately high availability and performance guarantee will still depend on the level of fundamental resource provisioning. Network slicing will though gain currency not just for cost saving through consolidation around a common infrastructure, but also for delivering other benefits or aspects of 5G. These include security as we discuss separately this week.