Network slicing technology has large parts of the wireless networking community excited, and it’s easy to understand why: It can deliver multiple logical networks over a single physical infrastructure and give each network the unique set of characteristics required to meet specific user needs.
The idea of re-using scarce radio-frequency spectrum to create more value for providers that own the infrastructure, in addition to the ability to offer over-the-top services beyond mere connectivity, has the telecom giants interested, particularly those deploying 5G cellular services.
What is network slicing?
Network slicing is a layer ofvirtualisation applied to wireless network services. Just as servers in the cloud or containers are virtual constructs, not simply physical servers, a network slice its own logical network carved out of a larger physical network via the automated allocation of bandwidth, QoS rules and other network functions.
The actual process of doing this varies by the company offering the technology, but it’s essentially an outgrowth of software-defined networking and network function virtualisation technologies, coupled with automation in order to dynamically manage multiple slices of a large network.
A network slice contains the same control plane, user plane and access network interfaces as any other network, it simply portions out parts of them as necessary to support multiple virtual networks. Using this technique, a single network can be subdivided logically into a host of different networks, each with capabilities and features appropriate to their use.
Spectrum is a finite commodity, and conventional QoS technology simply doesn’t do enough to maximise its use. During peak demand, large portions of spectrum may remain unused, while other portions might be uncomfortably crowded.
For example, a base station near a large public venue might be swamped with voice, video and app traffic while an event is in progress. At the same time, bandwidth the operator has set aside for machine-to-machine use, like smart meter reading or smart home services, remains mostly inactive.
With network slicing virtual networks could be reconfigured and re-provisioned largely on the fly to automatically shift bandwidth and other resources from the machine-to-machine use to the needs of the event goers, alleviating the crunch and providing better service.
Much of the conversation around network slicing centers on 5G because slicing is a key reason that that 5G will be able to provide new functionality carriers have advertised.
Telecoms are eager to offer services beyond simple connectivity to business users, and the ability to create individualised, virtual networks on the fly is central to that concept. Here are the three types of network slice currently contemplated by 5G proponents.
Enhanced mobile broadband
Mobile broadband is traditional data-to-your-cellphone service. Network slicing can support enhanced mobile broadband (eMBB), which is aimed at maximising the data rate while supporting moderate reliability and packet-error rates of about 10-3, according to an IEEE working paper on the subject.
A network slice devoted to eMBB would be provisioned with those constraints in mind, letting the system select and manage clear channels for use in eMBB. Massive machine-to-machine connections
Massive machine-to-machine connections
One implementation of the Internet of Things (IoT) is deploying large numbers of data-gathering devices that need to phone home only intermittently to provide small bits of data—massive machine-to-machine connections (mMTC). Rather than low latency and high bandwidth, mMTC requires the network to support huge numbers of low-bandwidth connections simultaneously.
The idea behind mMTC is to provide a maximal arrival rate for any given band of radio frequency, but allowing a packet-error rate around 10-1. That is much higher than the 10-3 target rate for eMBB but is perfectly acceptable given the small size of the messages being sent.
Ultra-reliable, low-latency communications
Many of the more novel wireless services made possible with network slicing fall into the category of ultra-reliable, low-latency communications (URLLC). These include support for critical infrastructure, medical services like remote surgery, connected vehicles, and real-time process coordination in fields like manufacturing, according to the Federal Communications Commission.
The IEEE says this type of transmission is intermittent, like mMTC, but its use of comparatively small blocks of data and strict latency requirements mean that a much higher level of reliability is required, on the order of a packet error rate of 10-5.
While the technology itself promises to be a fairly direct upgrade over current-generation methods of QoS and traffic shaping, the radio access networking (RAN) technology that enables its deployment remains relatively immature, and the standards governing it have yet to be finalised.
Even prominent wireless equipment vendors like Ericsson have only recently begun to deploy slicing capability in their radio-access-network technology, meaning that the hardware to implement it is still in the process of reaching the market.
What’s more, all the vendors involved are still producing pre-standard equipment.
The 3rd Generation Partnership Project (3GPP), an alliance of seven telecommunication standards-development bodies, is working toward standards, but has yet to finalise them for parts of the stack that are critical to implementing network slicing at scale. Multiple-input, multiple-output (MIMO)—sending and receiving multiple signals over the same channel at the same time—enhancements to vehicular wireless connectivity, and RAN slicing itself are all part 3GPP’s work and are expected to remain in the pipeline until at least mid-2022.