Challenges of Kubernetes for composable disaggregated computing in IOWN Global Forum

Innovative Optical and Wireless Network(IOWN) Global Forum who made MoU with Linux Foundation to collaborate for the Future Smart Connected World is now under development for an infrastructure of composable disaggregated computing in data centric manner so called DCI (Data Centric Infrastructure). IOWN DCI helps to enable Cyber Physical Systems(CPS) and AI Integrated Communications(AIC) across multi-sites over IOWN's Open All-Photonics Network(APN). IOWN DCI architecture concept has a capability to compose Logical Service Node with PCIe device pool such as GPU, DPU/IPU/Open Programmable Infrastructure(OPI), etc, to enable a composable radio access network, a composable multi-access edge computing, a composable AI inference ingestion, etc.
We are facing a challenge of Kubernetes for dynamic scale up/down in IOWN DCI, while enabling scale out/in.
In this session, we will discuss current gap that needs Kubernetes Device Resource Allocation(DRA) enhancement for GPU and DPU/IPU/OPI based on IOWN DCI PoC such as CPS and AIC, and we will elaborate Kubernetes' possibility in composable disaggregated computing. KubeCon attendees can come together in this session to discuss composable disaggregated computing interworking with Kubernetes and share their thoughts and ideas.

This talk describes one of the future server systems as proposed in the IOWN DCI concept and shows the benefits of such a new architecture. It also introduces emerging Composable Disaggregated Infrastructure(CDI) technology and how the CDI contributes to the IOWN DCI. You also know how the OPI project works in it.
Moreover, this talk helps you to understand what challenges we face to satisfy the requirements of IOWN DCI. Especially, when balancing programmable devices such as GPU and DPU with CDI.
In the end this talk shows how we try to cope with these challenges and Kubernetes DRA can be a key function for this solution.