Hardware Management

The OCP Hardware Management Project defines a small, stable set of tools for remotely managing machines at scale. Its specification incorporates existing tools and best practices for remote machine management. The project’s activities fall in five areas: manageability interface and capabilities, the baseboard management controller, rack management firmware, the management controller module, and hardware fault management.

Hardware Management

Get involved

Steering Committee Representative

John Leung

Project Leads

Subscribe to mailing list

About This Project

Scale computing requires a small and stable set of tools to remotely manage machines. The Hardware Management specification incorporates a set of existing tools and best practices and leverages existing tools for remote machine management.

Scope

The activities of the Hardware Management Project fall in the following areas:

The activities of the Hardware Management Project fall in the following areas:

  • Manageability Interface & Capabilities
  • Baseboard Management Controller
  • Rack Management Firmware
  • Management Controller Module
  • Hardware Fault Management

Resources

Manageability Interface & Capabilities

  • Standard manageability interfaces, which include the following capabilities: resource discovery; configuration and update; telemetry; control; asynchronous eventing and notification; and composition
  • Baseline manageability capabilities across the Server, Storage, Networking, Telco, Data Center Facility, and Rack & Power projects
  • Provide guidance and feedback to other Open Compute projects extending the baseline manageability capabilities
  • Overseeing the OCP’s repository for OCP Profiles

Baseboard Management Controller

  • Baseboard management controller (BMC)
  • Out-of-band manageability interface to the BMC
  • In-band (system or host interface) manageability interface to the BMC

Rack Management Firmware

  • A rack manager manages OCP platforms installed on one or more racks, and exposes that manageability to the rest of the data center
  • The rack manager is the combination of open-source rack management firmware executing on a contributed rack management controller
  • The rack management firmware will expose an interface which conforms to an OCP-prescribed interface
  • The activities for this area are in the purview of the OpenRMC-DM Sub-Project

Management Controller Module

  • The management controller module separates the common platform management hardware and places it behind a connector
  • The module is specified by its connector pin-out and may have one or more form factors
  • The activities for this area are in the purview of the Hardware Management Module Sub-Project

Hardware Fault Management

  • Requirements for hardware fault management
  • Standardizing system behavior under hardware failures
  • Define key baseline requirements of managing hardware errors to achieve target service level
  • Provide reference and guidance on system hardware failure management
  • The activities for this area are in the purview of the Hardware Fault Management Sub-Project

Regular Project Calls

This project meets on the third Tuesday of every month at 11am ET

Call Calendar

These meeting are recorded via audio and video. By participating you consent that these recordings may be made publicly available. Any presentation materials, proposals and meeting minutes are published on the respective project’s wiki page and are open to the public in accordance to OCP’s Bylaws and IP Policy. This can be found at http://www.opencompute.org/about/ocp-policies/. If you have any questions please contact OCP.

Hardware Management calendar

Related Details

Contributions

All project contributions

Sub-Projects

Fleetscale Memory Fault Management

Fleetscale Memory Fault Management (FMFM) is a Workstream within the OCP Hardware Management Project focused on standardizing memory fault management at fleet scale. It is proposing a vendor-agnostic architecture for memory error handling, with modular inputs from different hardware vendors, standard APIs between modules, and defined outputs such as failure cause, health information, and RAS actions. It is also standardizing memory error telemetry for better fleet-scale RAS management, troubleshooting, and FRU replacement policies, and coordinating with the broader OCP community to ensure a path for the architecture. The specification is developed on GitHub.

GPU Management

GPU Management is a Workstream under the OCP Hardware Management Project, which defines a small, stable set of tools for remotely managing machines at scale. Details about the workstream’s mission, scope, leadership, and ways to get involved will be added in the near future.

Hardware Fault Management

The Hardware Fault Management Sub-Project operates under the direction of the OCP Hardware Management Project. It defines requirements for hardware fault management and works to standardize system behavior under hardware failures, so operators can rely on consistent, predictable responses across OCP platforms. The sub-project sets key baseline requirements for managing hardware errors to achieve target service levels, and provides reference material and guidance on system hardware failure management for the wider Open Compute community.

Manageability Profile

The Manageability Profile Workstream is an open workstream of the OCP Hardware Management Project that publishes OCP manageability profiles for the Redfish interface. Products seeking OCP recognition must meet the Supplier Checklist's required profiles, shown by Redfish Interop Validator output. When the Incubation Committee approves a Usage Guide it also approves the matching profile, which is then released at the top level of the HWMgmt-OCP-Profiles repository on GitHub. The workstream collaborates with the Server Project, the OpenRMC-DM Sub-Project, the Cooling Environments HWLC workstream, and the Sustainability Project, and its profiles span baseline, server, rack, service-baseline, NIC, GPU, power shelf, rack PDU, liquid cooling, and sustainability management domains.

OpenRMC-DM

The OpenRMC-DM Sub-Project, under the direction of the OCP Hardware Management Project Group, creates designs and specifications that enable interoperable manageability for OCP rack managers. Its deliverables include a specification for the interface the rack manager exposes to the rest of the data center, requirements for the interface between the rack manager and on-platform components, and an open-source implementation of the rack manager controller software. The sub-project specifies the northbound and southbound interfaces of the rack management controller and oversees OCP's Rack-Manager source repository. Its Device Manager incubation adds a unified way to identify and query devices, monitor early warning signs, and manage software and firmware.

RAS API

The RAS API Workstream is a semi-private (CLA) workstream of the OCP Hardware Management Project. Without an industry standard for Reliability, Availability, and Serviceability (RAS) management, each hardware vendor’s solution is bespoke and RAS feature adoption is slowed. The workstream is drafting a specification for a common RAS abstraction between software and hardware that is vendor and platform agnostic, lets in-band and out-of-band connectivity co-exist, handles error storms and sampling in its logging, and is generic enough for open-source drivers while leaving room for vendor differentiation. Revision 0.8 is the current baseline.

Scalable Cloud Infrastructure Management

The SCIM workstream produces documents on architectures, protocols, and data models for carrying hardware management information across data center networks; analyzes existing communication mechanisms and engages standards bodies to improve them at scale; examines management challenges from new technologies such as AI/ML, RDMA, PCIe/CXL, DDR, and quantum computing; and develops predictive models linking hardware behavior to data center financial performance.

System Debug

The System Debug Workstream is an open workstream of the OCP Hardware Management Project addressing the diagnosis and debugging of data center components and nodes across their lifecycle. It produces architectural documentation defining the interfaces, protocols, APIs, and data models for communicating diagnostic and debug information; analyzes existing architectures and standards-body approaches; addresses scalability across the lifecycle, including accessible ports, diagnostic capabilities, and security and privacy requirements; and coordinates with other workstreams on shared constructs. Its work-in-progress specification is developed on GitHub.