Unified orchestration and lifecycle automation for end-to-end 5G deployment

Unified orchestration and lifecycle automation for end-to-end 5G deployment

Operators are currently modernizing their network for 5G deployments. Important elements in these transformations are the disaggregation of hardware and software. This transformation is seen in both the 5G Core (5GC) and Radio Access Network (RAN) architecture. Furthermore, the decoupling of hardware and software opens up the ability to support multi-vendor networks. SDN and NFV emerged as innovative technology for service providers to build their next generation networks.

3GPP specifications defined the 5GC network as service-based architecture, where each element in the core network is a virtual network function and there are service-based interfaces (SBI) to talk to each other. This architecture is suited for cloud-native micro-serviced based deployments. There is a huge transformation of Network Functions (NF) to cloud-native world. Kubernetes, open-source container orchestration platform, is the de-facto standard that operators are adopting for deploying their 5G workloads.

5G RAN and Open-RAN

RAN deployments were mainly driven by few vendors with proprietary implementation and RAN composition was typically purchased from a single entity during 3G/4G deployments. The lack of interoperability between RAN vendors forced operators to stick to the same vendor. According to 3GPP 5G standards, the RAN (gNodeB) components are split into the Central Unit (CU) and the Distributed Unit (DU) for flexibility in deployment. Open RAN (O-RAN) ecosystem extended this architecture to more open interfaces and disaggregation of hardware and software components. There are new interfaces introduced for interactive service management and orchestration framework and RAN Intelligent Controller (RIC). Furthermore, it provides mechanisms that analyze network conditions allowing applications to feed-back into the RIC’s decision making.

The O-RAN architecture is not just about splitting the RAN components. It includes an orchestration layer, for applications running on Consumer of The Shelf (COTS) hardware and introduces standard interfaces (O1 and A1) between orchestration layer and virtualization layer. This also offers operators a flexibility to orchestrate any multi-vendor CU and DU workloads in Edge sites or radio sites, on any x86 hardware infrastructure.

Infrastructure Automation and Orchestration

Operators are no longer buying proprietary hardware from vendors to build their networks, but building their data centers to host the applications using a multi-vendor ecosystem. 5G deployments typically span across multiple data centers including core, edge and far edge data centers. The edge and far edge (radio) sites are deployed at scale and are geographically distributed across the country. It is nearly impossible to manually onboard 1000s of servers across different sites. Thus, there is a need for automating and orchestrating the end-to-end lifecycle of infrastructure at scale. This includes installing the OS, configuring and updating bare-metal COTS servers, configuring networking and storage, installing Kubernetes clusters, onboarding NFs and Network Service (NS) lifecycle management. There are no industry standards defined around infrastructure automation (compute, storage and network resource), but this is a mandatory requirement for operators to build their network at a faster pace. At 5G scale involves numerous, error-prone, manual operations, just to ready the servers for deployment. There is a need to fine tune the edge servers with multiple settings at os and kernel to meet the throughput and latency requirements for CU, DU and UPF workloads. Automated bare metal server readiness will significantly reduce costs and reduce deployment times from days to minutes.

ETSI MANO architecture

When operators started adopting virtual network functions in their network, they saw the need for standards to automate and orchestrate that facilitated interoperability. European Telecommunications Standards Institute (ETSI) came up with Network Function Virtualization (NFV) Management and Orchestration (MANO) architecture for mobile networks and edge applications. ETSI MANO specifies different roles and interfaces involved in orchestration and automation of network functions on a virtualized infrastructure. Operators use this reference architecture to build their MANO framework for their 4G/5G deployments. This includes how to manage NFVI resources and VNF resources from the orchestrator and enable services on the network functions. MANO architecture started with automating and orchestrating virtual network functions (VNFs) and now extended the same to container network functions (CNFs).

The Problem statement (Operator’s challenge):

The architectural shift in the 5G network clearly has the need for automation and orchestration in the 5GC, RAN and infrastructure (data center or cloud). The operators have the flexibility to build a multi-vendor network, where they can pick and choose best of breed for each network function. The challenge that operators typically face, when it comes to orchestrating, automating and managing multi-vendor applications is choosing a Kubernetes platform that meets all of the needs of their application. Each NF vendor might have their own orchestration, automation and Kubernetes platforms. This creates a challenge for operators to manage multiple tools and interoperability between them. Also, operators have the flexibility of choosing different x86 hardwares from hardware vendors and each might have their own automation tools to configure and manage the server settings. There is a need for a single orchestration and lifecycle automation layer, with a single Kubernetes platform, managing everything, 5GC, RAN, supporting applications, appliances and the underlying x86 hardware infrastructure. The operators need a single pane of glass to operate, manage and visualize their 5G network. Also, they need to onboard the servers and network functions with zero touch provisioning.

Robin Multi Data Center Automation Platform (MDCAP)

Robin MDCAP is a multi data center, multi-cloud automation tool that can orchestrate deployment of all the components of 5G networks including bare metal servers, Kubernetes clusters, RAN and Core network functions. Before onboarding any network function or installing Kubernetes cluster, a server must be updated and configured. Robin MDCAP performs full bare-metal life cycle management and can verify, install, upgrade, configure and bootstrap your server infrastructure. These are not simple scripts, they are contextually aware workflows with numerous user defined checks points and forks, that guide your installation to its desired conclusion. Transform a server, without configuration or operating system, via the Baseboard Management Controller (BMC), Ethernet or serial connection. Monitor and manage readiness:

  • Basic Input Output System (BIOS) and BMC configurations
  • NIC, Solid State Drives (SSD), Field Programmable ASICs (FPGA), non-Volatile Memory express (NVMe), Redundant Array Of Independent Disks (RAID), firmware upgrades
  • OS installation, Drivers, Services and Software Packages

Robin MDCAP orchestrates and manages the lifecycle of any workflow including, bare-metal provisioning, cloud platform instantiation, network functions, network services and methods of procedures (MOPs), all of which can be auto-triggered through a policy engine. MDCAP’s automated workflows support Cloud Native NFs (CNFs), Virtual Network Functions (VNFs) and Physical Network Functions (PNFs) simultaneously. MDCAP can be deployed in central DC and can manage all the remote DC servers, kubernetes clusters and applications like RAN Centralized Unit (CU), Distributed Unit (DU) and User Plane Function (UPF), 5G Core (AMF, SMF, PCF) that span across different regions and some of them can co-locate with radio sites. Robin MDCAP software comes with a built-in workflow manager, where the network admin can build different workflows for the above-mentioned use cases. The workflow can be installing an OS on a bare-metal server and installing a Robin cloud-native platform in High Availability (HA) mode or instantiating a NF on a cluster. There are a set of predefined workflows in MDCAP workflow factory. Customers can define custom workflows inside MDCAP workflow studio according to their needs. The workflow will support anything from bare metal server configuration to all the way to network service instantiation and life cycle management.

Robin CNP (Telco Carrier grade Kubernetes platform)

Robin CNP (Cloud-native platform) is the carrier grade Kubernetes platform that is fine-tuned to run 5G Core and RAN applications across edge, central data centers. Robin CNP starts with open standards, upstream Kubernetes without change and adds telco grade CNI that supports network acceleration technologies including Single Root I/O Virtualization (SR-IOV), Data Plane Development Kit (DPDK) and persistent IP and Multus plugin support. The Container Network Interface (CNI) also provides overlay networking (Calico) and Open vSwitch (OVS). The next addition is an enhanced Container Storage Interface (CSI) storage layer that includes unique application aware data services including, snapshots, backup and clone and complex service level management capabilities.

Robin also added an enhanced application workflow manager that interfaces with Kubernetes scheduler and supports complex placement policies like NUMA-Aware CPU pinning and complex affinity and anti-affinity rules and hugepages support. The Scheduler also added a Container Runtime Interface (CRI) multiplexer that allows seamless deployment of VMs (VNFs) on containers, enabling efficient resource sharing and removing operations silos. Robin also added enhanced observability capabilities and application aware automation capabilities.

Robin carrier grade networking and advanced workload placement can meet all the typical requirements of 5G applications which require high-throughput, low latency and scalability. Robin allows you to model your resource needs and dependencies with a simple graphical declarative interface. Once modeled, in seconds, Robin applications seek out and configure the resources for you. Furthermore the modeled policy is pinned to your application and can automatically reconfigure over stops, starts, heals and migrations, for the lifecycle of the application. Some of the key features offered for 5G network functions include:

  • CPU isolation and NUMA aware placement of network-functions
  • Fast path with enhanced throughput – SR-IOV with DPDK
  • Multiple types of network interfaces (Management, FCAPS, Data traffic)
  • NIC redundancy using bonding feature
  • Persistent IP and static IP allocation for the NF using robin IPAM, over the entire lifecycle of the application
  • Multiple gateway support for network functions
  • Network functions with dual Stack support (IPv4/IPv6)
  • A network with mixed VNF / CNF workloads
  • Bare metal performance w/o virtualization overhead

The industry is currently transitioning from virtual machines to cloud-native solutions and will be doing so for many years. There are some network functions that are from the 3G/4G world which may not be containerized soon. Robin provides a platform to run both CNFs and VNFs simultaneously. Using Robin CNP platform operators will be able to build their network with a mix of VNFs and CNFs and they don’t need to wait for vendor roadmaps to complete their application containerization strategy, hence faster to market.

Additionally, CNP supports Multi-tenancy and role-based access on any cluster so that operators can bring in multiple vendors and allocate resources required for each vendor application. CNP has built in monitoring tools to monitor the cluster with open-source tools such as Grafana and Prometheus. Operators can integrate with their own monitoring tools to monitor the cluster and network functions running on it.

Conclusion

Robin enables you to operate 5G network deployments with unmatched lifecycle simplicity, performance, scale, and advanced workload placement. By utilizing Robin MDCAP with its bare metal to services orchestration and the industry’s most advanced cloud platform, Robin CNP enables:

  • Significant reduction in deployment time, from days/hours to minutes
  • Enabler of HW/SW disaggregation, driving CapEx reductions up to 40%, Hyper-automation can reduce OpEx up to 50%
  • Robin infrastructure was over 30% faster compared to OpenStack based VNF platform on the same hardware due to the efficiencies of containerization
  • Automated and resilient life-cycle management: Full stack life-cycle management of the HW platform, SW platform, Cloud platform and CNF/VNF services chains
  • Simplified deployment: One-click onboarding, with an easy-to-use declarative model that scales, heals and migrates using autoconfigured, service-pinned policies
  • Flexible, High-performing: Industry leading cloud platform, designed for 5G applications: A cloud-native, Kubernetes-based platform that supports flexible networking options, VNFs/CNFs, application aware storage and advanced multi-parameter, multi-cluster, workload placement
  • Network wide monitoring, analytics and closed loop automation
  • Single pane of glass view to the application and infrastructure
  • Zero touch provisioning and onboarding of servers to the network
  • Increase in network availability, thanks to self-healing and multi-application placement
  • Very low footprint for edge and far edge RAN deployments

Share with: