Chris Wright, Chief Technology Officer at Red Hat, conducted a keynote session and elaborated upon the open source possibility around IOWN technology development, and shared Red Hat’s contribution to the IOWN GF.
In the IOWN GF technology working group, we are contributing to three task forces: The Data-Centric Infrastructure Task Force, IOWN for Mobile Network Task Force, and the Reference Implementation Model Task Force. We joined the discussion in each TaskForce last year from our perspective based on OpenShift/Kubernetes and RHEL upstream possibilities, and delivered the reference documents summarizing minimum viable versions of IOWN technology in 2021 together with IOWN GF members in each Task Force.
Data-centric Infrastructure is designed to enable service providers with the tools needed to build and flexibly place data pipelines, including data processing and storage functions, and to dynamically select data transfer and network protocols for each. Figure 1 provides an example of a Logical Service Node providing Tenant Workloads for overlay solutions. It also shows an enhanced Logical Service Node integrating telco specific accelerators and a vendor’s embedded proprietary code, such as for 5G Radio Access Network. They can run on Kubernetes worker nodes as composable compute in a disaggregated infrastructure with Open APN (All Photonics Network) providing lossless and low latency infrastructure.
Having the ability to isolate tenant workloads and Infrastructure functions on the same data-centric infrastructure system is one of the principal advantages of this architecture. Different types of workloads can run effectively as composable compute, such as the Infrastructure control/management plane on the domain specific hardware like a DPU/IPU (denoted Function Dedicated Network unit in IOWN GF team).
Cleanly compartmentalizing the hardware accelerated platform addresses the mismatch among abstraction layers that compounds the ongoing complexity of software stacks. Figure 1. Red Hat technology contribution in IOWN Global Forum
Compartmentalizations challenge in disaggregated infrastructure
Traditional disaggregated computing architecture brings complexity to resource management. It is challenging to realize Cyber Physical System Use Case and AI integrated Communication Use Case because there are multiple owners of the resources across Customer edge, Telco Edge, Telco Core and Cloud in the end-to-end computing infrastructure for Fixed and Mobile Convergence 6G.
HyperShift is an open source project that can help to solve the issue of complex management for disaggregated computing resources as well as standalone COTS resources with domain specific hardware. HyperShift is middleware that hosts OpenShift control planes at scale. It helps reduce cost and time to provision, as well as allowing portability across clouds with strong separation between management and workloads. It is paving the programmatic way shown in Figure 2 to help solve complex management issues in disaggregated infrastructures like IOWN’s.
HyperShift also aids in solving the excess overhead problem of having to deploy a controller node for each control plane. A new feature under upstream development allows for separating Kubernetes worker pools and Kubernetes controllers - running the different control planes as managed services on a consolidated management cluster.
The Tenant workload cluster(s) and the infrastructure cluster are regular OpenShift instances, and thus can run ecosystem applications consumed from the marketplace, creating a convenient way to deploy based on GitOps principles. Essentially this means moving to higher-level APIs and abstractions at the service level.
It matches the hardware architecture to the concept of orchestrating encapsulated services that containers and Kubernetes have generalized for application software. We could interpret this as the ‘containerization’ of domain-specific hardware. Efficiencies such as these could drive the industry to disaggregate CPU-centric systems and shift to a composable compute model, where compute capabilities are software-defined.
GitOps controllers, such as Red Hat Advanced Cluster Management (ACM), the community project version of ACM and Open Cluster Management (now a CNCF project), would be able to manage compartmentalized resources for tenant workloads as well as infrastructure functions using HyperShift. Figure 2. Architecture compartmetarization
Last blog post shared our technology possibilities through IOWN panel discussion with NTT and Intel held at Open5Gcon2021 . This blog post introduced more specifics on how Red Hat technology is contributing to OPEN Data-Centric Infrastructure development. Presently, we are contributing to the IOWN GF technology working group and the PoC consultation Task Force accelerating PoC for evaluating IOWN technology with our recommended open source software. Futher report of IOWN PoC based on RHEL/OpenShift technology will be psoted at Red Hat blog site in near future.
In the meantime, we invite you to keep up with the latest telecommunication ecosystem news.
About the author
Hidetsugu (Hyde) Sugiyama is a Chief Architect at Red Hat, focused on the digital service provider and telecom sector in Japan. Sugiyama has been with Red Hat for over eight years, working on software-defined networking, network functions virtualization, edge computing solutions development, and joint go-to-market with technology partners and R&D customers. He has 34 years of experience in the information and communications technology industry.