IT Systems Workshop
The IT Systems Workshop is a community workshop of the OCP Open Systems for AI Strategic Initiative, held virtually on July 29, 2025 as the "Server Designs and Management for AI" session of the initiative's 2025 workshop series in the DC IT Infrastructure work area. It brought together practitioners to share knowledge and vision on AI server designs and management, with sessions on the Mt. Diablo server scope, high-efficiency storage cooling, AI infrastructure management through open standards, the Open Cluster Designs initiative, CXL tiered memory and MRDIMM, memory-efficient RAG pipelines, and DC-MHS. The agenda, slides, and recording are available on the event page.
Get involved
About This Workstream
The IT Systems Workshop is a community workshop of the OCP Open Systems for AI Strategic Initiative, held virtually on July 29, 2025 as part of the initiative’s 2025 workshop series. It is the “Server Designs and Management for AI” workshop in the DC IT Infrastructure work area.
Open Systems for AI is the umbrella Strategic Initiative that coordinates the OCP Community’s work across its Projects and other Initiatives toward an open ecosystem for AI data center infrastructure. Its scope covers guiding and highlighting key AI infrastructure contributions and facilitating workshops and OCP Foundation alliances, organized into three work areas: DC Physical Infrastructure, DC IT Infrastructure, and Systems Management.
Data center IT infrastructure refers to the components and systems that make up the AI clusters supporting AI workloads — the silicon, IT hardware, and interconnect. Building on the initiative’s earlier Physical Infrastructure workshops, this workshop brought together practitioners to share their knowledge and vision on AI server designs and management and to allow for a melding of forward thinking in this space.
Sessions
- Mt. Diablo Server Scope — Christina Peabody (Google), Mingchun Xu (Meta)
- High Efficiency Mass Storage Cooling Innovation — Mohamed El-Batal (Seagate)
- AI Infrastructure Management: Challenges and Mitigations through Open Standards — Chitrak Gupta and Sudan Ayanam (AMI)
- Open Cluster Designs Strategic Initiative — Lowell Lamb (Broadcom), Loren Staley (Celestica)
- Next-Gen AI/HPC Server Performance with CXL 3.1 Tiered Memory and MRDIMM — David Zhuo (Montage Technology)
- Towards Memory-Efficient RAG Pipelines with CXL Technology — Arun George (Samsung), Siamak Tavallaei (Independent)
- DC-MHS — Mike Gregoire (Dell)
Workshop materials
The full agenda with presentation slides is on the event page; the session was recorded.
Scope
Topics covered in this workshop:
AI server designs — the Mt. Diablo server scope and the Data Center Modular Hardware System (DC-MHS)
Storage and cooling — high-efficiency cooling for mass storage in AI systems
AI infrastructure management — challenges and mitigations through open management standards
Open cluster designs — an introduction to the Open Cluster Designs for AI Strategic Initiative
Memory for AI/HPC — CXL 3.1 tiered memory, MRDIMM, and memory-efficient RAG pipelines