The Innovative Optical and Wireless Network (IOWN) Global Forum is developing next generation data-centric infrastructure over all photonics networks (APN), to help realize attractive use cases for AI Integrated Communications(AIC) and Cyber-Physical Systems (CPS). IOWN Global Forum's approach will have a strong influence on telecom, media and entertainment and other industries in the green transformation era.
This summer, IOWN Global Forum released four initial proof-of-concept(PoC) reference documents that Red Hat, together with NTT and other members, has contributed to. These are:
These PoC reference documents refer to the 1st edition of DCI (Data-Centric Infrastructure) Functional Architecture that was published in early 2022. Here is how Red Hat OpenShift supports these IOWN PoCs.
The intention of this PoC is to demonstrate that RDMA over APN removes the obstacles of long-range communication from the forum’s use cases. The following figure illustrates PoC based on the IOWN reference implementation model for the area management security use case in CPS for massive sensor data aggregation and ingestion, with RDMA over APN technology to increase data velocity for AI-inference. While the current Open Data Hub framework does not include a solution to increase data velocity at the ingestion stage in the data pipeline, this IOWN reference implementation model will help AI/ML users increase end-to-end data processing speed by leveraging IOWN technology. OpenShift supports GPUdirect RDMA and RoCEv2 (RDMA over Converged Ethernet v2) by using data processing units (DPU) or RDMA-capable network interface card (NIC) that this PoC needs.
Figure 1: RDMA over APN in the context of Area management Security use-case
The intention of this PoC is to demonstrate the benefits and viability of APN as a fronthaul solution. The recognition of mobile fronthaul over APN as a viable and promising solution is expected to encourage the participation of service providers around the world and to further promote IOWN-related technologies. The following figure illustrates the network’s capability of supporting high availability services and elastic load balancing. This term of elastic load balancing means the active and dynamic steering of radio access network (RAN) components including the radio unit (RU) connections to a set of (virtual) distributed units (DUs) based on actual load. APN can dynamically switch wavelength paths, allowing them to redirect computing resources at the destination according to traffic volume and other factors. Thus, APN can be a solution that allows service providers to dynamically allocate and deallocate DU computing resources. In PoC step 2, the forum focuses the vDU deployment on top of the logical service node that may be running either on the entire COTS(Commercial Off-The-Shelf) server or on the entire DPU/IPU(Infrastructure Processing Unit). DU hosts that are used as a logical service node running OpenShift could run multiple vDU cloud-native network functions (CNFs) from our CNF certified partners.
Figure 2: vDU mobile fronthaul network over APN
Last is the PoC for Data-Centric Infrastructure-as-a-Service (DCIaaS). DCIaaS is the foundation for the IOWN technology to deploy the IOWN data hub (IDH) and the IOWN mobile network (IMN) for 5G RAN, multi-access edge computing (MEC) and 6G (under development), in addition to CPS and AIC deployment. The DCIaaS PoC is to demonstrate a concept of logical service node (see figure 3) that is composed dynamically from allocatable hardware device resource pools (CPU, GPU, DPU/IPU, etc.) to realize each use case in CPS and AIC, and to demonstrate the advantages of the open all photonics network across customer sites, regional edge sites and core sites. OpenShift can compose a Kubernetes base logical service node on an x86 CPU bare metal host (or virtualization environment). In addition to an x86 CPU base logical service node, OpenShift can run on the DPU as an enhanced logical service node in this DCIaaS PoC.
Figure 3: OpenShift base Logical Service Nodes in Data-Centric Infrastructure
Collaboration with Open Programmable Infrastructure project
Through the activity in the forum, it was realized that a fully-programmable open infrastructure model across software stacks and DPU/IPU-like hardware devices is needed to develop network, security or storage functions. The Open Programmable Infrastructure (OPI) project aims to explore and expand the concept of programmable infrastructure using community-driven, standards-based, open ecosystems for next generation architectures and frameworks based on DPU/IPU-like technologies.
This summer, the OPI was established by the founding members (including Red Hat) as a Linux Foundation project. The ongoing collaboration between IOWN Global Forum and the OPI project could help both communities reach mutual goals, with Red Hat and other ecosystem partners acting as catalysts for technological innovation and exploration of new use cases.