A Busy person's guide to Network Slicing

Network Slicing is one of the many promises made by 5G Technology. In this post, I'll try to provide some grounding information on this most fascinating concept

Jun 30, 2024

sliced orange fruit on white surface — Photo by Hưng Nguyễn on Unsplash

What is a Network Slice ?

Let’s start with the basic question - What is a network slice? . Well, to answer that question, we will look into how 3GPP (The 3rd Generation Partnership Project - an umbrella standards organization that deals with world-wide standards on telecommunication technologies) defines Network Slicing.

As per 3GPP,

A network slice is a logical network that provides specific network capabilities and network characteristics, supporting various service properties for network slice customers.

Well, that’s quite mouthful isn’t it? Lets try to break it down.

A telecom network consists of 3 main parts as follows:

Access Network which is consumer facing, where we connect our gadgets to. This typically includes Public WiFI access points, Cellular base stations, Broadband Optical Fibers & Cables etc
Data Network which is the producer facing, where the internet resides. Or in other words, applications such as NetFlix, Google, Amazon Prime and everything that makes up the internet resides.
Core Network which connects the Access to Data networks and enables a Communication Service Provider (CSP) aka Telecom companies to monitor, charge users & maintain overall network.

Each of this parts, can be considered as a sub-network (aka subnet) of a overall network. Each CSP has one ore more networks to cater to their subscribers. Sometimes, a CSP may rent / lease a portion of their network or their subnet to other CSPs for interoperability and other reasons.

Now lets understand what goes into each subnet. Every subnet performs a specific job in enabling connectivity to the end users. The systems running inside each of this subnets are termed as Network Function(s) (NF). Some typical examples of NFs could be network switches, routers, load balances, DNS servers, AAA servers etc. These NFs work with each other to enable a virtual pipe or a tunnel through which, an end user’s data session can be established between the User Equipments (UEs) (aka your gadgets like mobile phone, Laptops etc) to the Data Network (aka internet)

How a telecom network enables internet to end users

Now consider a case where a disaster like earthquake or hurricane stuck and suddenly everyone is scrambling to contact their loved ones. At the same time Govt & Emergency services also want to use the same network to coordinate the rescue and law & order activities. But a network, like any other physical infrastructure is scarce and cannot be multiplied indefinitely. As one can imagine, the moment where there is more demand to consume anything than its supply, naturally the notion of regulation, prioritization takes birth.

In prior 5G world (i.e 4G / 3G ) there is a concept called Quality of Service (QoS). QoS is basically a way to implement the above prioritization so that, special users get special treatment. What that means is, if an emergency responder and a common public tries to establish a connection on the same time and the network is congested, the network would prioritize the emergency responder’s traffic as opposed to common public. They do that by tagging the traffic the originates in UE as special and then in the Core network, non-special packets will be sent on a slow path, whereas special packets will be sent on a fast path to reach the internet.

You can visualize QoS as a scene normally you would see in airports where, passengers with TSA Pre-check gets whisked away into the security checkpoint while we commoners have to go through long lines of airport security.

So what happens in 5G world ?

Well in 5G, 3GPP introduced lot of new innovations and ground breaking concepts to modernize telecom network. While discussing every innovation in 5G would too much exhaustive, some things that are worth mentioning are Control plane User Plane disaggregation (commonly known as CUPS), Network Functions Virtualization / Network Virtualization (NFV), Software Defined Networking (SDN), Cloud Native network functions (CNFs).

What this effectively means is that, the entire telecom network can be modelled as a Computer in itself aka Network as a Compute model. Just like how we virtualize our laptop by creating virtual machines such that in a single laptop hardware, you can have 1 or more logical machines (VMs), similarly with a single telecom network infrastructure (i.e access + core + data), many logical networks can be created. That’s the broad idea.

Why do we even need a network slice?

Network Slices provides many benefits to a CSP than to a normal consumer of a network. Infact, for an average consumer this concept is totally transparent unless marketed by the telecom companies. Users doesn’t care how the pipe is formed and all they care about is 2 things

Speed at which they can connect to internet
Duration i.e How long can they consume (whatever they consume) at a reasonable speed.

Hence all these 5G jargons are basically for the service provider and the surrounding network enabler ecosystems. So, let’s try to understand from a service provider perspective, why they would need a network slice.

Network Optimization - Simply put, network slices can help telcos (short form of telecom companies) to maximize their ROIs on their network infrastructure. In many countries, telcos have moral & social responsibilities to ensure they provide uninterrupted network connectivity to masses. They should also ensure network stays alive especially during global and national disasters & tragic events. Often communication technologies are treated as national resources and Govt sometimes lease the resources for the telcos to license and do business. In return these telcos have to ensure they provide necessary value in ensuring connectivity to different sects. Hence from a service provider perspective, they need to effectively use their existing network to serve different agendas. For such agendas, instead of creating a dedicated network, network slices provide the ability to repurpose their existing network to different needs without compromising on the quality of services that they need to fulfill legally.
E2E Visibility - As per 3GPP definition, A network slice is a logical network which is end to end (E2E). What is E2E here? Recall that a telecom network is made of different parts i.e subnets. When a network slice gets created, it contains all parts that make up a complete network. Hence this slice can perform independently from other slices. That’s where this term E2E comes in. By creating network slices, Network Operators have complete visibility over all slices which by definition are E2E.
Network Elasticity - Just like how an elastic rubber band expands and contracts based on the size of the things it is holding in, Network Slices provides ability to create, modify, destroy several logical networks on demand based on the needs. This is effectively a boon for Telcos. Using a single set of network infrastructure, they can easily scale up / scale down networks based on the load condition and necessity. For eg, if there is an indoor event happening and more people concentrate on a specific place, a CSP can simply create a slice on demand to serve that particular location. This way, resources from other locations are temporarily routed to this location to cater the load and once the event is finished, that network can be shutdown thereby freeing those resources. Other example could be, on a disaster place, a temporary secure network can be spun up on-demand for the law enforcement and it can be shutdown once the recovery efforts are completed.

2-minute detail on how network slices work ?

Okay now that we have a fair understanding (I assume 😇) on what are slices and why do we need them, let us try to understand how they work. This is a quick 2-min watered down version for busy persons.

A Network Slice has a lifecycle of its own. Following are different stages in which a network slice can be in:

Planning - It all starts with the Telcos deciding to design a slice for some reasons. It could be due to network optimization needs for their internal operations or a dedicated slice they wanna build for a different customer, sector or other administrative needs or on demand situation where they need to stand up a network temporarily etc. Once the decision to create a slice happens, a network slice is formally designed. Network Slice Design encompasses different details like Slice Composition (what makes up a network slice), Deployment Location, Slice Capacity, Lifetime Duration etc.
Provisioning - Once the slice has been designed, it needs to be instantiated or deployed. Consider this as a case where you have a machine ready, but you need to power it on and maybe configure a little to make it work according to your needs. In telco world, this step is often referred as Provisioning.
Operating - Once the slice is up & running, it needs to be constantly monitored and reconfigured as per changing network needs. This is as good as someone operating the machine. For eg, you operate the car while driving on the road and you constantly use the car’s control to dynamically adapt its behavior as per the traffic conditions.
Decommissioning - Finally once the telcos determine that the network slice has served its purpose, they dismantle the slice thereby freeing up the resources allocated for that slice. This is the final stage of the slice in its lifecycle

When a slice is deployed, a Network Slice Instance (NSI) is said to have been created. You can understand NSI as a objective form of a slice living in the network, such that this slice can be uniquely identified with some form of ID. For eg, we as person exist, but our passports or driving licenses identify who we say we are legally. Similarly, NSIs gives an identity to slices.

When one or more NSIs are deployed in the network and when a user equipment needs to utilize them, there should be some way to discover the existing slices right? Its like a google search for NSIs. That’s where some special network functions comes into play. You see, when an operator deploys a network slice, the core network will have to make a note of this NSI. The network function called Network Slice Selection Function (NSSF) will register all such slices and will serve as a database of NSIs. So when a user wants to play a cloud game, by consuming their subscription, they make an indirect request to the network to consume the underlying slice that powers this cloud gaming subscription. The Access network will search an instance of this slice type that powers the cloud gaming service. The Core network will discover a NSI using NSSF and then activate this user’s data type within that NSI. Voila the user’s traffic tunnel is now placed on a highway directly to the Gaming server thereby bypassing all loops and intersections that would otherwise would become the path of normal traffic flow.

Slice types

When a network slice is designed & deployed as part of an ongoing planning activity, sometimes these network slices are termed as Static Network Slice or simply static slices. Whereas, if a network slice is designed and deployed on demand based on some trigger condition being met, it is termed as Dynamic Network Slice or simply dynamic slices. 3GPP in collaboration with GSMA (a non-profit organization made of worldwide GSM Operators) has defined or in other words standardized Static Slices for industry’s benefit. They are as follows:

eMBB Slice - Enhanced Mobile BroadBand - this will be the underlying slice type that powers our day to day internet.

uRLLC Slice - Ultra Reliable Low Latency Communications - for advanced sci-fi technologies like remote surgeries, metaverse gaming etc, you need lightning speed. This slice type powers such things (atleast that’s the intention).

mMTC Slice - Massive Machine Type Communications - anything that deals with machine to machine communication without human in the loop i.e your alexa talking to your smart fridge or ring door bell. This doesn’t need huge data transfer like we humans needs to process. All such usecases are powered by this slice type.

In addition to these standardized slices, a CSP can also create a custom slice that fits for their needs as well.

Again for end users like you and me, it doesn’t matter which slice type we consume. CSPs will use these slice types internally to standardize their operations and provide a billable plan or service that we commoners will simply subscribe, pay and consume it from our devices.