Industry Thought Leadership

Network of the Future - Dynamically Sliced, Differentially Priced

November, 2018
Manish Singh
VP of Network Services Strategy & Marketing

Tech Mahindra

Digital Transformation is reshaping almost every industry, heralding unprecedented opportunities and challenges for each vertical. Digital continues to shape the new marketplace as brick-and-mortar gives way to e-commerce, structured industries yield to uberization and content goes over the top. Disrupt or be disrupted, is the mantra of this Digital marketplace. And 5G is about hypercharge this Digital Transformation.

5G will be the transformational tipping point that will deliver unprecedented productivity gains, redefine competitive advantages and accelerate global market reach reshaping every industry vertical (including agriculture, automotive, healthcare, manufacturing, energy, transport, gaming, media & entertainment etc.). Businesses will reexamine their P&L with a 5G lens. Network will become the differentiator that will redefine business processes, foster business innovation and pioneer new production, distribution and consumption models. Those who prepare well for 5G have much to gain and will be the disruptors. Others run the risk of being disrupted.

5G Use Cases
To meet the diverse Digital requirements of different industry verticals, 5G use-cases are broadly classified into three buckets: eMBB (Enhanced Mobile Broadband), MIoT (Massive IoT) and uRLLC (Ultra-Reliable Low Latency Communication). Each with different set of requirements for the network:

  • eMBB: requires ultra-high bandwidth, low latency to deliver enhanced user experience on mobile devices and to enable new immersive experiences like AR/VR. Lowering cost-per-bit is a major driver to continually meet the latent consumer demand for more and more data.
  • MIoT: requires low-bandwidth and even low data rates to connect massive number of sensors, but must support extended battery life of sensors that will power Smart Homes, Smart Stadiums, Smart Industries, Smart Cities and more.
  • uRLLC: requires ultra-reliable, ultra-low latency, high bandwidth to enable new applications like Autonomous Vehicles, Telemedicine, Precision Manufacturing Robots and more.

Following three network KPIs (Key Performance Indicators) succinctly capture the use cases’ diverging requirements:

  • eMBB requires peak 10Gbps and average 1Gbps data rates,
  • MIoT requires 1M devices/km2 connection density, and
  • uRLLC requires 1ms latency.

5G must solve delivering one network that can meet all these diverging requirements

5 Laws shaping 5G
5G has to solve for delivering one network that meets all the aforementioned diverging requirements. Following are the five laws that are fundamentally shaping the 5G network:

  1. LAW OF PHYSICS: Nothing travels faster than the speed of light (c ~ 3 x 108 m/s). Today, applications hosted in the Cloud typically deal with ~100ms network latency. 5G is targeting to enable ultra-low latency applications that need <10ms latency. The only way this network latency requirement can be realized is by bringing both the network and the end application closer to the UE. Edge Computing (MEC) and Control-User Plane Separation (CUPS) in 5G will be pivotal in enabling these ultra-low latency applications.
  2. MOORE’S LAW: Compute (actually transistors) density doubles every two years, creating ever hungrier devices/things and more processing power for the network infrastructure. Apple’s A11 processors have six ARM cores and Intel’s latest Xeon processors, Purley, have up to 28 x86 cores. Devices will exponentially demand more and more data. And 5G network too will greatly benefit with the growing infrastructure compute capacity, enabling sophisticated, real-time complex algorithms. MU-MIMO, Beamforming, Beam tracking etc. will all be beneficiaries of Moore’s Law in 5G, paving way for leveraging smart antenna techniques to deploy higher spectrum bands. Increases in compute capacity will enable processing 100MHz wide channel bandwidth in 5G and in handling 1Gbps average data rate. If partitioned right, Telco Cloud can even beat Moore’s Law.
  3. LAW OF THE LAND: Regulatory frameworks across different countries will significantly shape 5G. Net neutrality is back at the forefront and will pave the way for how 5G networks will be dynamically sliced and differentially priced. Spectrum licensing regimes will shape how 5G networks will be rolled out. Licensed, unlicensed and even shared spectrum regimes will shape both CSP 5G networks as well as Private 5G Networks. Furthermore, licensing regimes need to be harmonized; for ex: 60GHz today is unlicensed in US, but licensed in India. Policymakers will have to collaborate to harmonize spectrum bands to preserve the economies of scale and facilitate cross-border roaming for mobile users. Lastly, as applications gets hosted at the network edge, data privacy regulations like GDPR will become increasingly relevant for 5G networks.
  4. LAWS OF ECONOMICS: Production precedes consumption; production has costs while consumption is the ultimate goal. 5G network rollout will incur significant CapEx and OpEx spend from CSPs. Which applications, which industry verticals, which use-cases will drive new monetization opportunities remains uncertain in the short term. Dynamic Network Slicing in 5G networks will provide the essential flexibility to CSPs to deal with uncertainty and balance network Demand-Supply in this brave new Digital world of fail-fast and scale-fast.
  5. LAWS OF THE CLOUD: Nothing is shaping 5G more than the principles of Cloud. Cloud has clearly established the power of centralization with unprecedented economies of scale. Scale matters and best of breed wins in the world of Cloud. On demand, self-service/DIY is the established Cloud operating model. Disaggregation of Network Functions (NF) and SDN/NFV are the bedrock foundations of the Network of the Future. Both Core and Access Networks will be virtualized in 5G and will form the essential building blocks for Network Slicing. 5G gives CSPs the unique opportunity to break traditional vendor lock-in and build their networks with best-of-breed solutions. Stovepiped back office functions will yield to Dynamic Orchestration. Self-Service/DIY, On-Demand operating model will holistically require a new approach to traditional BSS/OSS back office functions.

5G Infrastructure: A Distributed Telco Cloud
Networks are inherently hierarchical and distributed in nature. 5G Infrastructure will be a Distributed Telco Cloud built by leveraging merchant silicon, COTS hardware and Cloud OS. Network Functions (NFs) themselves will be dis-aggregated, virtualized, will be truly cloud native and deployed as workloads on the Telco Cloud. Decoupling NFs from the underlying infrastructure enables them to be orchestrated and dynamically placed at different locations of the Telco Cloud, thereby, enabling creation of logically independent network slices on a shared infrastructure. This in turn enables CSPs to create multiple logical network slices that best meet the varying requirements of bandwidth, connection density, latency, elasticity and cost. The distributed infrastructure will further enable CSPs to host NFs as well as 3rd party applications from different industry vertical in a way that best meet both the technical and business requirements. Edge Cloud in particular will bring compute and storage capabilities closer to the end device/thing (UE) and will be critical for meeting the requirements of ultra-low latency applications and for efficiently managing ultra-high bandwidth applications.

A three tier Distributed Telco Cloud Architecture can vary depending on CSP’s existing infrastructure, geographic footprint and end application requirements. Distributed Telco Cloud will enable CSPs to achieve the Cloud-like economies of scale and simultaneously balance stringent requirements like ultra-low latency. The Distributed Telco Cloud operating model will be: “Centralize what you can. Distribute only what you must”.

5G Network
5G is changing the network’s anatomy to enable CSPs to offer wide range of Digital Services. Historically, appliance centric networks were engineered to meet the requirements of limited set of services like voice, messaging, data and video. These networks however have proven to be inflexible to meet the diverse and varying requirements of Digital Services. As we move past the era of limited killer-apps to an era of diverse Digital Services, 5G is holistically redefining the network end-to-end, including both the Access and the Core Network.

5G AN: Access Network
To meet the diverging needs of eMBB, MIoT and uRLLC use cases, 5G AN (Access Network) will be dis-aggregated and virtualized. To meet the ultra-high throughput and ultra-low latency requirements, 5G NR (New Radio) will operate in diverse spectrum bands spanning across low, mid and high spectrum bands. To deliver average 1Gbps data rates and to solve for the cell edge problem, 5G AN will employ advanced phased antenna array technologies including MU-MIMO and Beamforming.

In the world of Cloud, best-of-breed has already won. With 5G, CSPs now have the opportunity to dis-aggregate and virtualize the RAN (vRAN) and pave the way for best-of-breed 5G AN solution. Leveraging the vibrant eco-system of vBBU, RRH, AAS, SON and COTS white-box suppliers, CSPs can drive margin stacking out of their supply chain and break the large OEM’s vendor lock-in on their RAN. Benefits of vRAN go well beyond the equipment cost, as it greatly simplifies/reduces the RAN site deployment cost and significantly reduces OpEx by driving energy and cooling costs down at the Base Station (BS) sites. vRAN further provides added benefits of energy management, dynamic network scaling and high availability, all of which will be key requirements for 5G.

5G splits the gNodeB architecture into Centralized Unit (CU), Distributed Unit (DU) and the Radio Unit (RU). The architecture offers various options for splitting the gNB stack across real-time and non-real time functions of the RAN. The split choice would depend on a number of factors including underlying transport network, spectrum channel bandwidth, fronthaul bandwidth requirements, connection density and latency tolerance. Above all, vRAN will enable end-to-end network slicing in 5G network.

5G NR (New Radio) will enable operators to tap into mmWave spectrum bands, above 24GHz, to create fat OTA (over the air) pipes. Ample spectrum available in these mmWave bands, facilitate CSPs to create network that will deliver ultra-high throughput and ultra-low latency. However, use of these higher spectrum bands have their own set of challenges including high propagation loss, required line-of-sight (LoS) and high susceptibility to signal blocking. Challenges span across both indoor and outdoor environments as typically these signals cannot even penetrate walls and can easily be blocked by foliage or even moving vehicles in urban environment. Deploying in mmWave bands necessitates sites to be in close proximity. Poles and other street furniture provide ideal deployment sites as these bands would be best suited for dense urban population areas that need massive capacity boost. Advanced antenna techniques including beamforming and beam tracking will be employed for mobilizing mmWave in non-line-of-sight environments. In addition to mmWave, CSPs will also employ diverse spectrum assets below 6GHz and even below 1GHz to provide ubiquitous 5G coverage. Balancing traffic demand against deployment costs will govern how CSPs acquire and deploy their spectrum assets to create adequate coverage and capacity for their 5G network nationwide.

5GC: Core Network
5G Core will be truly cloud-native, deployed on a Distributed Telco Cloud infrastructure, delivering unprecedented Service Agility. SDN and NFV are the foundational pillars of the 5G Core Network. For the very first time, 3GPP has specified Service Based Architecture for 5G Core, in addition to the traditional Reference Point Architecture (used in 2G, 3G, 4G and now 5G). Service Based Architecture holds the key to rapidly enabling wide range of Digital Services with varying requirements from different industry verticals.

Existing mobile core networks have only one set of logical control functions and this in turn has artificially limited the service set that CSPs can offer to their customers. If a new service introduction required modifying an existing network function or introducing new network function(s), then CSPs had to undertake the complex task of integrating and retesting their entire core network end-to-end and would typically take anywhere from six to eighteen months – this will no longer works in the Digital Services era. This architectural rigidness has fundamentally inhibited CSPs to rapidly offer new services and efficiently seize market opportunities. In sharp contrast, web-scalers have demonstrated rapid service introduction and elastic scaling capabilities. Service Based Architecture brings that cloud agility to the 5G Core Network. Different Network Functions can be quickly composed and orchestrated into a new end-to-end service making it easier to add, modify or remove NFs from service chains and even quickly realize new service chains.

Just like microservices interact to create complex cloud applications, 5G Core will be composed of self-contained reusable software modules of network function services. This modularity enables dynamic composition of different network cores on a per slice basis, with varying capabilities, that best meet the end service requirements. Also like cloud application, 5G core network functions will mostly be stateless decoupling the compute and storage resources which in turn enables faster scale-out, flexible NF placement and new software enabled redundancy models for high availability and geo-redundancy. Minimizing state will be a key challenge for 5G NFs and key is to keep stateful services to a minimum.

Cloud native, Service Based 5G Core lends itself very well for Network Slicing by allowing reuse of network function services and even rapid network function customization, as needed, across slices. In addition to the traditional core network functions like authentication, session management and mobility management, 5G Core introduces new functions like Network Slice Selection Function (NSSF), NF Repository Function (NRF) and Network Exposure Function (NEF) to support cloud native Core and Network Slicing. NRF maintains NF repository enabling registration and discovery of could native NFs for cloud-like operations. NEF is like an API Gateway, allowing external users to define and enforce policies per their end application requirements. NEF is particularly important to enable Digital Services for diverse industry verticals and enable cloud like Self-Service/DIY operational models. NSSF enables Network Slicing by assisting selection of the right Network Slice instance, in a multi-sliced core, that best meet service specific requirements.

An important aspect of the cloud native 5G Core is the separation of Control and User Plane functions. Web-scalers have used similar architectures to efficiently scale their cloud applications with centralized control and distributed processing to efficiently scale sessions and transactions. Control and User Plane Separation (CUPS) in 5G Core will enable independent scaling and evolution of control plane and user plane functions. Architecturally this will enable networks to efficiently scale and respond to data tsunamis and/or signaling storms, as the case maybe. Furthermore, this will enable flexible deployment of user plane functions at regional or edge clouds to optimally meet the high throughput and low latency requirements of the end application.
5G Core will be Access agnostic. This will enable CSPs to deploy a converged core network which will integrate both 3GPP ANs (like 5G NR, 4G LTE) as well as non-3GPP Access Network (like WiFi).

Network Slicing
Network of the Future will be dynamically sliced and differentially priced.

Network Slicing will enable CSPs to create different logical networks, on a shared Telco Cloud infrastructure, with varying capabilities that best meet the varying requirements of different applications. Throughput, latency, security, mobility requirements can all vary from application to application. Furthermore, economic models across eMBB, MIoT and uRLLC will greatly vary and in turn will put varying demands on the network.

To enable cloud-like On-Demand, Self-Service/DIY models for different industry verticals, Network Slicing will pave the way for CSPs to offer differentiated services at differential price points. CSP can offer network slices with varying throughput, latency, mobility, security etc. to the end user and can even offer varying levels of granular control for the slice.

Harnessing the power of Virtualized RAN (vRAN) and cloud native Core, 5G network can dynamically allocate resources and generate network topologies with the right set of NFs (Network Functions) that best meet the end service requirement. Resources can be dedicated or shared across slices. A slice itself can support one or many services.

In a nutshell, Network Slicing can be summed up as the network adapting itself in software to deliver optimal network topology capability and resources that best meet the end service requirements. To illustrate the adapting capabilities of a 5G network, let’s take three different use cases with three different service requirements:

  1. Virtual Reality (VR): This use case can be classified under eMBB service which requires high throughput and low latency. To optimally meet these requirements, the network slice would take advantage of CUPS and place Core User Plane (UP) function at the local DC, closer to the end device to optimize latency. Further, 3rd party Gaming Applications and VR Acceleration engines can be hosted and integrated to further optimize service delivery. To meet the high throughput requirements, 5G AN would employ advanced antenna techniques like MU-MIMO and beamforming.
  2. Autonomous Car: This use case requires uRLLC service to meet the stringent ultra-low latency and high throughput requirements. To meet these stringent requirements, the slice would deploy both real-time and non-real-time functions of the 5G AN as well as the 5G Core UP right at the Central Office (CO) Edge Cloud; bringing them closest to the end device. Network would leverage multi-mode AN connectivity to meet high reliability and high throughput requirements. Further, 3rd party autonomous car V2X server can be hosted at the CO Telco Cloud to reduce the application latency even further.
  3. IoT Sensors: This use case requires MIoT service to meet the low bandwidth and long battery life requirements. These sensors would mostly be static (low mobility) and largely generate small volume of uplink data periodically. The economics of this market are going to be ultra-low ARPU. Aligned with the principle of “Centralize what you can. Distribute only what you must”, this network slice would deploy all the network functions in the centralized Hyperscale Telco Cloud. CSPs can also host 3rd party IoT servers on their infrastructure and even go up the value chain by offering Managed Digital Services.
    End-to-End Network Function Dis-aggregation and Virtualization (of both Access and Core Networks) are the fundamental building blocks of this adaptable and sliceable 5G Network. Ergo Network Slicing necessitates utmost diligence in selecting best-of-breed NFs to achieve highest levels of Service Agility.

5G Rollout
Barring an unforeseen Greenfield operator, 5G will be rolled out in phases. 3GPP standardized Non-Standalone (NSA) mode in December 2017, and in June 2018 3GPP released the standards for Standalone mode (SA) as well.

Most of the 5G trials and early rollouts will be deployed in Non-Standalone mode leveraging 4G LTE Core and RAN for ubiquitous coverage and 5G NR selectively deployed to boost capacity and data rates. This provides the fastest path for CSPs to launch 5G commercial services. Non-Standalone mode is also currently being used by CSPs to study and characterize RF Propagation characteristics of mmWave spectrum band.

Eventually, to reap the full benefits of 5G, CSPs will migrate to Standalone mode. Enabling wide range of Digital Services will require CSPs to deploy 5G Core Network and automate end-to-end Network Slicing. Progressive CSPs have already initiated deploying their Telco Cloud, LTE vRAN and Virtualizing their 4G EPC. Further, they are also deploying CUPS in their 4G EPC, paving the way for 5G Network and simultaneously minimizing stranded investments.