BoF: IOWN Data Centric Infrastructure as a Service PoC Discussion 🔗 **Source:** [BoF: IOWN Data Centric Infrastructure as a Service PoC Discussion](https://www.youtube.com/watch?v=YVjM5Vy4KLo) > [!note]+ Document Notes > tags: [IOWN, composable-infrastructure, OPRA, DCI, photonic-networking] > author: Linux Foundation > source: YouTube > reference: https://www.youtube.com/watch?v=YVjM5Vy4KLo > date created: 2024-12-19 14:30 > > # 🏷️ Title: BoF: IOWN Data Centric Infrastructure as a Service PoC Discussion > <!-- deep-read:sections start --> ## 🧰 Action Inbox - [H1] Develop composable disaggregated infrastructure POC — combine CDI with accelerator technologies #infrastructure - [H2] Present CDI prototypes at next KubeCon — show effectiveness of solution #presentation - [H3] Standardize communication between DRA and CDI — address orchestration challenges #standardization ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [IOWN Global Forum](https://iowngf.org/) - [Open Programmable Infrastructure Project](https://opiproject.org/) ## 📝 Summary - IOWN (Innovative Optical and Wireless Network) Global Forum presented their data-centric infrastructure proof-of-concept collaboration with Linux Foundation - NTT, Fujitsu, and NEL are developing composable disaggregated infrastructure (CDI) for next-generation computing - The initiative aims for 100-200x performance improvements through photonic networking and disaggregated computing - Key challenge involves unified management of diverse accelerator technologies and dynamic resource allocation - Bottom line: IOWN DCI represents a paradigm shift from server-based to component-level resource pooling ## 📖 Main Content IOWN Global Forum, established in 2020 with 135 members, is developing next-generation ICT infrastructure targeting quantum-level improvements in bandwidth, power efficiency, and latency. The forum collaborates with Linux Foundation on three key projects: Open Programmable Infrastructure (OPI), LF Health, and LF Networking. The core innovation involves shifting from traditional server architectures to fully disaggregated computing infrastructure where components (CPU, GPU, DPU, IPU) are connected via high-speed photonic networks and composed dynamically based on workload requirements. This approach enables purpose-built logical service nodes rather than fixed server configurations. NTT demonstrated two scaling patterns: Pattern A uses OpenShift on x86 CPU to expand multiple GPUs, while Pattern B enables micro-shift inside DPU/GPU accelerators for scale-out. The composable disaggregated infrastructure (CDI) allows dynamic attachment/detachment of PCI devices, optimizing resource utilization and power consumption. ## 🗒️ Notes #### ⏱️ [00:00](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=0) — [H1] Introduction to IOWN Data Centric Infrastructure - Key idea: IOWN Global Forum collaborating with Linux Foundation on next-gen infrastructure - Evidence: "MOU with Linux Foundation to collaborate with three projects: OPI, LFH, and LF Networking" #### ⏱️ [02:30](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=150) — [H2] IOWN Global Forum Overview - Key idea: 135-member forum targeting 100-200x performance improvements - Evidence: "Target performance will be 100 to 200 better than conventional technologies" #### ⏱️ [05:45](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=345) — [H2] All-Photonic Network Architecture - Key idea: Shift from electrical switching to direct optical connection - Evidence: "Networking of IOWN will fully leverage optical technologies, shifting from repeat switching to direct connection through APN" #### ⏱️ [07:20](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=440) — [H2] Disaggregated Computing Infrastructure - Key idea: Component-level disaggregation connected via photonic networks - Evidence: "Component devices composing conventional servers are disaggregated and connected via high-speed photonic network" #### ⏱️ [10:15](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=615) — [H2] Real-time Video Analysis POC - Key idea: Demonstrating geographical resource distribution with lower latency - Evidence: "Demonstrate efficient utilization of geographically distributed resources with lower latency and power consumption" #### ⏱️ [12:00](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=720) — [H3] Two Server Configuration Patterns - Key idea: x86+DPU based and converter device-based AI inference - Evidence: "Two different server configurations: x86 plus DPU based and converter device based for AI inference" #### ⏱️ [15:30](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=930) — [H2] Dynamic Scaling Requirements - Key idea: Infrastructure must adapt to changing business and workload demands - Evidence: "Key word for DCI is dynamicity - type and number of accelerators can vary depending on various factors" #### ⏱️ [18:45](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=1125) — [H3] OpenShift Integration Patterns - Key idea: Two deployment patterns for container orchestration - Evidence: "Pattern A: OpenShift on x86 CPU to expand multiple GPUs; Pattern B: micro-shift inside DPU and GPU accelerators" #### ⏱️ [22:10](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=1330) — [H2] OPI Project Collaboration - Key idea: Enabling Linux OS inside DPU/IPU for intelligent infrastructure services - Evidence: "OPI targeting to enable Linux OS inside DPU and IPU to increase intelligence of infrastructure services" #### ⏱️ [25:40](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=1540) — [H2] Composable Disaggregated Infrastructure Concept - Key idea: Software-defined composition of custom servers from resource pools - Evidence: "CDI disaggregates existing servers into separate components and composes custom-made servers by software definition" #### ⏱️ [28:20](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=1700) — [H3] CDI Resource Pooling Mechanism - Key idea: Components connected via PCI/CXL/photonic switches with software control - Evidence: "All components connected to PCI or CXL or photonic switches and CDI management software controls switches" #### ⏱️ [31:00](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=1860) — [H3] CDI Performance Benefits - Key idea: Dynamic device attachment for workload optimization - Evidence: "When system used for AI training, CDI attaches GPUs; for inference, CDI detaches GPUs and saves power" #### ⏱️ [35:15](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=2115) — [H2] Current CDI Challenges - Key idea: Need standardized communication between orchestrators and CDI - Evidence: "Two challenges: specify standard communication between DRA and CDI, and interworking mechanism for programmable devices" #### ⏱️ [38:30](https://www.youtube.com/watch?v=YVjM5Vy4KLo?t=2310) — [H3] Future CXL Integration - Key idea: Next-year plans include CXL switch deployment - Evidence: "Next year we will use CXL switch providing CXL memory and CPU pool capabilities" ## 📚 References - [IOWN Global Forum Official Site](https://iowngf.org/) - [Open Programmable Infrastructure Project](https://opiproject.org/) - [Linux Foundation Networking](https://lfnetworking.org/) ## 🗣️ Paraphrase - IOWN represents a fundamental shift from traditional networking to all-photonic infrastructure enabling direct optical connections - Composable disaggregated infrastructure allows dynamic assembly of computing resources from component pools rather than fixed servers - The collaboration between IOWN Global Forum and Linux Foundation OPI project addresses unified management of diverse accelerator technologies ## 📰 Newsletter The IOWN Global Forum is pioneering next-generation data center infrastructure that could revolutionize how we think about computing resources. Instead of buying fixed servers, imagine dynamically composing exactly the CPU, GPU, and memory you need for each workload from a shared component pool. This composable disaggregated infrastructure (CDI) approach, demonstrated through collaboration with the Linux Foundation's Open Programmable Infrastructure project, promises 100-200x performance improvements while reducing power consumption. The key innovation lies in connecting components via photonic networks rather than traditional electrical switches. For WordPress hosting providers, this technology could enable unprecedented resource efficiency - automatically scaling GPU resources for image processing during peak hours, then reallocating them for video transcoding overnight. The challenge now involves standardizing management interfaces across diverse accelerator vendors. Resources: [IOWN Global Forum](https://iowngf.org/), [OPI Project](https://opiproject.org/), [LF Networking](https://lfnetworking.org/) ## ✅ Action Items #### DevRel - Developer advocacy: Present CDI concepts at cloud-native conferences, demonstrate POC implementations - Education: Create workshops on composable infrastructure principles, photonic networking basics - Developer experience: Develop APIs for CDI resource allocation, improve orchestration documentation - Developer marketing: Highlight 100-200x performance improvements, create case study content - Community management: Foster collaboration between IOWN members and open source communities #### OSPO - Policy & Governance: Establish contribution guidelines for IOWN-Linux Foundation collaboration - License Compliance: Review photonic networking stack licenses, ensure compatibility - Security Management: Assess security implications of disaggregated infrastructure - Software Inventory: Track CDI component dependencies, maintain SBOM for photonic stack - Developer Support: Provide training on OPI project integration, tooling for CDI development - Community Engagement: Support IOWN Global Forum membership, fund upstream development - Strategic Alignment: Align CDI roadmap with organizational cloud strategy, track performance KPIs #### Community - Maintainers: Support OPI project roadmap development, governance template creation - Contributors: Enable onboarding for photonic networking contributions, first-issue programs - Sponsors: Transparent budgets for IOWN collaboration, impact reporting on CDI adoption - Ecosystem: Coordinate between IOWN, LF Networking, and cloud provider initiatives ## 🧪 Hypotheses - Composable disaggregated infrastructure will reduce data center operational costs by 30-50% through improved resource utilization - All-photonic networks can achieve sub-millisecond latency for geographically distributed computing resources - Dynamic resource allocation using Kubernetes DRA can effectively manage CDI component attachment/detachment - Unified management interfaces can abstract vendor-specific differences in DPU/IPU implementations ### 📊 Methodology - Performance benchmarking comparing traditional servers vs. CDI configurations across various workloads - Power consumption analysis measuring energy efficiency gains from dynamic component scaling - Latency measurements of photonic vs. electrical networking under different traffic patterns ### ❓ Questions - How will CDI handle failure scenarios when components are dynamically attached across photonic networks? - What standardization efforts are needed to ensure interoperability between different vendors' photonic switches? - How can Kubernetes resource scheduling be extended to optimize CDI component allocation? ### 🔮 Future Research - Integration pathways between CDI and existing container orchestration platforms - Security implications of disaggregated infrastructure with photonic interconnects - Economic models for CDI adoption in enterprise and cloud provider environments ### 💡 Implications - Practice: Data center operators need new skills in photonic networking and disaggregated system management - Policy: Industry standards required for CDI component interfaces and management protocols - Security: New attack vectors emerge from disaggregated infrastructure requiring updated security models - Education: Training programs needed for next-generation infrastructure technologies - Funding: Significant investment required for photonic network infrastructure deployment ### 🧠 Learning Styles Assessment Reading level estimate: 12th Grade #### 👀 Visual learners - Present: Architecture diagrams showing CDI components, photonic network topology - Improve: Animation showing dynamic resource composition, power consumption visualizations #### 🔊 Aural learners - Present: Video presentation with speaker explanations - Improve: Podcast format discussion, audio summaries of technical concepts #### 🗣️ Verbal learners - Present: Technical definitions, acronym explanations, analogies to traditional infrastructure - Improve: Storytelling approach to CDI benefits, comparison narratives #### 🧩 Kinesthetic learners - Present: POC demonstrations, hands-on system configurations - Improve: Interactive CDI simulators, step-by-step deployment guides #### 🧠 Logical learners - Present: Performance metrics, logical architecture progression - Improve: Decision trees for CDI adoption, quantified benefit analysis #### 🧑‍🤝‍🧑 Social learners - Present: Collaborative forum discussions, industry partnership examples - Improve: Working group participation opportunities, peer review processes #### 🧘‍♀️ Solitary learners - Present: Technical documentation, self-study materials - Improve: Individual exploration environments, personal reflection prompts #### 🌿 Naturalistic learners - Present: Power efficiency improvements, sustainability benefits - Improve: Environmental impact comparisons, lifecycle analysis ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure, technical term definitions - Improve: Enhanced alt text for architecture diagrams, ARIA labels for complex visualizations #### Keyboard navigation - Present: Standard video controls, navigable content structure - Improve: Keyboard shortcuts for video chapters, skip links for technical sections #### Contrast and readability - Present: Standard presentation format - Improve: High contrast mode for technical diagrams, larger text options #### Text resizing and reflow - Present: Responsive content design - Improve: Ensure technical diagrams scale properly, maintain readability at 200% zoom #### Media and captions - Present: Video presentation with audio - Improve: Accurate captions with technical terminology, downloadable transcripts #### Forms and interactions - Present: Basic YouTube interface - Improve: Accessible feedback forms for technical questions #### Standards reference - Ensure video player meets WCAG 2.2 AA standards for media accessibility ### 🌍 DEIB Assessment #### Cultural breadth - Present: Japanese technology leadership perspective, global forum collaboration - Improve: Include perspectives from other regions, translate key concepts #### Inclusive language - Present: Technical terminology with some explanations - Improve: Plain language summaries, avoid cultural assumptions about infrastructure knowledge #### Neurodiversity support - Present: Structured presentation format - Improve: Content chunking with clear sections, consistent terminology usage #### Representation - Present: Technical leadership from Asia-Pacific region - Improve: Diverse speaker representation, varied industry perspectives #### Psychological safety - Present: Professional technical presentation - Improve: Encourage questions from all experience levels, provide multiple learning pathways ### ✅ Logical Validity #### Strong reasoning - Deductive strength: Clear progression from infrastructure limitations to CDI solutions - Inductive strength: POC demonstrations support broader performance claims - Causal reasoning: Well-established connection between resource disaggregation and efficiency gains - Analogical strength: Effective <!-- deep-read:sections end -->