Skip to main content

Open Programmable Infrastructure

Open Programmable Infrastructure (OPI) is an open, vendor-neutral project that defines frameworks, APIs, and reference implementations for integrating and managing data processing units (DPUs) and other offload accelerators within modern data center and cloud infrastructure (infrastructure architecture).

  • Common, open framework for integrating DPUs, IPUs, and infrastructure accelerators into data center stacks (infrastructure architecture).
  • APIs and abstraction models for offloading networking, storage, and security services to accelerators (infrastructure offload).
  • Reference software stacks and architectures for DPU-based infrastructure services (reference implementation).
  • Enablement of multi-vendor, interoperable deployment of programmable infrastructure devices in clouds and enterprises (interoperability).
  • Collaboration forum for silicon vendors, system OEMs, cloud providers, and software projects around programmable infrastructure (industry collaboration).

More About Open Programmable Infrastructure

Open Programmable Infrastructure (OPI) addresses the integration and lifecycle management of data processing units and other programmable accelerators that offload infrastructure tasks such as networking, storage, and security from host CPUs (infrastructure offload). The project focuses on creating a common, open framework so enterprises and cloud providers can deploy and operate DPUs and related devices from multiple vendors using consistent abstractions and tooling.

OPI defines architectural models and APIs that describe how DPUs and infrastructure processing units connect into host systems, networks, and control planes (infrastructure architecture). This includes how infrastructure services such as virtual switching, routing, encryption, firewalling, and storage virtualization can execute on accelerators while remaining manageable through standard interfaces. The aim is to provide a uniform way for orchestration systems, operating systems, and management platforms to discover, configure, and monitor these devices.

The project provides reference software stacks and example implementations that illustrate how to build and deploy infrastructure services on DPUs (reference implementation). These artifacts serve as concrete patterns for integrating programmable infrastructure into existing data center environments, including cloud-native and container-based platforms. By publishing open models and code, OPI enables vendors and users to align around common building blocks rather than proprietary, device-specific integrations.

OPI is positioned for use in enterprise data centers, telecom environments, and hyperscale or private clouds that adopt DPUs or similar accelerators for offloading host Central Processing Unit (CPU) workloads (enterprise infrastructure). Typical use cases include virtualized networking, storage offload, encryption and security services, and observability or telemetry functions implemented directly on the Data Processing Unit (DPU). The project focuses on enabling these functions to be composed into higher-level infrastructure services that can be managed alongside traditional server and network resources.

The project operates as an open collaboration among hardware vendors, system integrators, cloud providers, and software communities (industry collaboration). Its work products, including specifications, APIs, and reference code, are designed to interoperate with existing data center ecosystems, enabling integration with orchestration frameworks, operating systems, and network or storage control planes where appropriate (ecosystem integration). In an enterprise technology directory, Open Programmable Infrastructure aligns with categories such as infrastructure architecture, hardware offload and accelerators, network and storage virtualization, and cloud infrastructure enablement.