Mark Gilmour outlines how 5G network slicing will allow MNOs to deliver the capacity, speed, latency and availability necessary to meet the differing needs of 5G use-cases and services.
Recently I was at an industry event in London where one of the speakers announced that 2017 was the year when the work to enable 5G will be needed. Certainly there is a lot happening across the industry as academia, standards bodies, forums, operators and vendor communities focus on trying to achieve the far reaching target specifications that have been well documented for 5G.
By the end of this year we should be very close to the first 3GPP 5G draft standard as release 15 is finalized, and as an industry we are putting pressure upon ourselves to be 5G ready. MNOs are pushing suppliers to validate that their networks are made 5G ready, and industry is still working on what that means exactly. No doubt 2017 will be a significant year for 5G.
One of those 5G target specifications is for 1ms latency on the radio interface. This is a significant development for mobile wireless networking and achieving this target specification opens a whole new world of use cases that previously were out of the reach of wireless connectivity.
Previously, applications sensitive to delay could not be deployed on a radio access network such as 3G HSPA or even LTE 4G networks, where the latency is measured in 10s of ms. Achieving 1ms on the air interface allows applications such as immersive gaming, augmented reality, and autonomous vehicles to be realised. The grey section of Figure 1 shows some examples of these previously not possible use cases.
As the air interface figuratively steps out of the way in terms of latency for these use cases, a greater emphasis is placed upon the network to become aware of the differing latency requirements. Therefore, network designers need to look at the placement of the BBU or CU (Centralized Unit in 5G architecture), and the placement of computing infrastructure to meet the latency budgets.
The simple answer would be to take the most latency sensitive application and design to meet that requirement. With network allocation targets of between a few 100 usecs to 10ms* across the broad spectrum of applications, this would require compute to be placed right at the edge of the network. This is simply not economically viable, and in some cases geographically possible, so a model which caters to different use case is required as illustrated in Figure 2.
Figure 2. Illustrative example of compute placement due to Latency/Use case requirements
5G networks will have to be adaptable, dynamic, and programmable from end-to-end using virtualised constructs. This is facilitated by implementing network slices for each supported use case with tailored performance invoked autonomously and programmatically. A fully flexible end-to-end 5G mobile network will allow MNOs to offer a far broader range of communication services built on existing, new, and yet to be created use cases that will ensure innovative and ongoing revenue streams beyond simple connectivity and capacity.
This flexible approach to the network architecture is required to cater for these different latency requirements - or more correctly different use case requirements - because latency is not only the performance differentiator for the applications. Use cases enabled by 5G may be differentiated by factors such as bandwidth, mobility, resiliency, security, availability and user attachment to name but a few. Therefore, the single ‘one size fits all’ construct of existing mobile networks becomes difficult to make a convincing argument for.
Network Slicing: Evolving the one size fits all construct
The end-to-end network needs to be separated, customized or sliced into discreet and isolated network partitions that allow the individual programmability to meet the differing needs of each use-case or service. 5G network slicing will allow MNOs to offer virtual networks over the same physical mobile network used by everyone else that will allow network performance per slice to be adjusted in terms of capacity, speed, latency, and availability.
Figure 3. NGMN Network Slicing Definition
The NGNM defines network slicing as:
“Multiple independent and dedicated virtual sub-networks created within the same mobile network infrastructure created to address mobile services having completely different requirements for latency, reliability, throughput, and mobility.”
The definition describes the components that comprise the network slice construct including the service instance, network slice instance, sub-network instance(s) and resource layer. In order to achieve the individual performance characteristics mentioned earlier, the network slice will need to dissect the whole end to end path. Therefore, 5G network slicing will surely be far more involved and challenging than traditional VPNs because it will involve the combination of many different types of existing and emerging networks technologies related to wireless, wireline, RAN, EPC, SDN, NFV, optical and smallcells. It will face far more demanding use cases, many of which haven’t even been dreamt up yet.
Regardless of the formidable technological challenges of network slicing, it’s seen by many as the best way to justify 5G investments, as it enables a wealth of new and exciting possibilities to not only justify investment but also to improve global communication in the future.
Mark Gilmour is Director of Portfolio Strategy for Mobile at Ciena - a network strategy and technology company.