According to the latest Market updates the global 5G active connections will surpass 580M in 2021 with astounding growth rate of 100% will reach 2.9B by 2025 , added in woes are machine and things that will add around 75B by 2025 .
Although these results are very encouraging it will raise challenges of its own kind not faced by mankind before requiring new architectures and principles to manage networks at scale . Issue like Data ,Privacy and Security is changing the Telco’s business narrative and more and more business now linking their business vision not such on Telco business but also on sustainability and responsibility .
The Flatter architectures and Cloud principles promised a great future by making it possible to Orchestrate and Automate networks and to use data to something not done before in any Telco generation and that is build Autonomous and Intelligent Networks by design solving some of the great challenges for 5G network scale and operational efficiency and this is topic i like to discuss today .
Orchestration and Automation for the Future World
Orchestration refers to the way the end user model, provision and manage the applications. The very nature of 5G which necessitates a distributed cloud and thousands of clusters it is vital we can handle all infrastructure in a software fashion that is friendly to use like drag and drop from intent point of view
From Telco perspective open and highly performant orchestration is the backbone for 5G Cloud infrastructure. Manual deployments of 5G services and connecting them using legacy approaches will be complex, error prone and not resource efficient .
Decoupling of Application and Infrastructure is a vision that orchestration solutions make it possible by using declarative API like YAML, TOSCA, Terraform today as it makes it possible to make Infrastructure irrelevant for the Application and hence to make Infrastructure totally Immutable that is provisioned using standard artefacts and templates. Declarative means an end user only defined “What” without specifying details of “How” . In fact 5G Cloud infrastructure is open and flexible in terms of How as it can use an extended set of tools to deploy it . It is by virtue of these characteristics that Telco’s target evolves from a manual to a Level4 autonomous networks of Future. As highlighted above the Telco for 5G and Edge applications will require some enhancements and that will require new frameworks today CNI plugins and CRD (Customer resource definitions) provided by different vendors for their offering made it possible to ensure all the Telco required enhancements can be deployed in a open cloud through open frameworks of Helm and concord .In addition, Orchestration will not only support vision towards software define Telco but also automated management of all the 5G infrastructure all the way from Physical servers/storage to the Application itself.
Network Slicing and B2B/B2B2X Models for Future
Network slicing is the segregation of one physical network in to a number of logical networks each serving varying use cases and business tenant that meets the desired SLA for different tenants .
To achieve this goal, Network Slicing needs to be designed from an E2E perspective, spanning over different technical domains (e.g. device, access network, core network, transport network and network management system). As example of reference architecture of a future network as shown below
However as today still there are a number of gaps which need more cross community collaboration, as an example the 3GPP SA5 resource model does includes modelling of the TN end-points it does however not include modelling for the 5G transport network itself, nor the RAN Furthermore, indication of whether a slice may share resources or not is indicated as part of the ServiceProfile. This indicates cooperation with other bodies, e.g. ETSI, as mentioned above, is needed. However the problem is that many other bodies define management function and interfaces regarding what and how they could allocate resources. Yet, there lacks of end-to-end view since transport and NFVI is not part of 3GPP. It is expected to specify management framework for SLA compliance and that is ongoing in SA5 with regards to RAN and Core. In addition, if resources are handled by vertical industry customers directly, further discussion will be needed. Based on our industry efforts we are bringing cohesion among following standardization organization for commercialization of 5G slicing
- 3GPP RAN
- 3GPP SA
- Broadband Forum – 5G Transport architecture
- IEEE 802 -Switched Ethernet networking and TSN
- MEF – Transport Services for Mobile Networks
- IETF – IP, MPLS SegRtg, EVPN, DetNet)
- OSM and ONAP
- ETSI NFV
The integration of automation and Telco DevOps for automating the end to end slices means E2E all the services can be provisioned in an agile manner from current 1Week to 1hours which is necessary to pace up with the innovation required in 5G era.
One of the typical issues with Slicing is that as tenant we need single Pane not just for services (GST or NEST) but also way how to connect them. Today frameworks like GANSO (GST and Network Slice Operator) is supporting industry to standardize on it .
AI/ML and Closed Loop for 5G Cloud Infrastructures
AI for Telecom has gained industry interests recently primarily driven by both wide deployments of 5G platforms which generates 4X more data compared to early generations alongside other global events like Covid-19 which necessitates a close loop operation avoiding the human. This initiative require not only orchestration but infact intelligent policy generation based on real time use and customer behaviors and will enable a SLA based offering for each 5G business tenant.
The use of ML/AI is still in initial phases of standardization, to ensure realization of a successful autonomous networks so the ML/AI should address following domains
- Automation and Policy
There is not just the technical side of ML/AI use in Telecom but a business side also. As we are well aware that many of NFV/SDN products in the market today that comes with native ML/AI functionality which are enabled not only in intent driven software level but also in chipset level one such example is Intel Atom , Intel 3rd gen Xeon processors with built in bfloat16.
support that reduces data required to build training models . However still Telco’s in 5G are trying to find sweet spot that will make business case of 5G positive. This is a fact that to build same coverage as 4G we need to pump 4 times more sites which means use of ML/AI for automated managed and use cases to optimize infrastructure is mandatory. In this context we also need to evaluate new business models for 5G to see “If 5G data can be monetized than Service can Free. From Infra view to Managed services view to vertical industry offering view”
In this context in 2021-2022 era I think Telco’s need to evaluate and commercialize following key cases for 5G ML/AI to speed up the deployment
- Life Cycle Managmeent of Infrastructure
- Automating Application and Infra Dependency
- Automatic output rule to Optimize NW specially RAN and Transport
- Advanced AI e.g build New Network Topology ,
- Work load placement , SLA analysis in case of PoP migration
The Telco operators should take active interest in following industry efforts to successfully use ML/AI in 5G Cloud Infrastructure
- ITU-T Focus group on ML for future NW (FM ML5G)
- ETSI enhanced network intelligence (ENI)
- O-RAN alliance for RIC (RAN intelligent controller)
Although the content should be enough for some but obviously for future networks the topic can not be considered complete unless we adress 5G in terms of security , public cloud integration in Telecom and Hybrid network Managmeent including evolution and migration with legacy networks . If that sounds interesting then keep following my blog as these shall be my topics of upcoming blogs .