Unlocking the Potential of Programmable Infrastructure: Introducing the OPI Project

Table of Contents

1. Introduction
2. Overview of the Open Programmable Infrastructure Project (OPI)
3. Working Group Leaders and Their Roles
4. About SNIA
5. Legal Disclaimer
6. Provisioning of DPU and IPO Devices
7. Inventory Collection from DPU and IPO Devices
8. Boot Sequencing
9. Lifecycle Management
10. Monitoring and Telemetry
11. Storage APIs
12. Networking APIs
13. Security APIs
14. AIML APIs
15. Simulation Environment and Developer Platform
16. How to Join the Project

Introduction

In this article, we will provide an in-depth overview of the Open Programmable Infrastructure Project (OPI) and its various working groups' efforts to develop standardized APIs and protocols for data processing units (DPU) and intelligent processing units (IPU). The OPI project aims to create an open ecosystem where multiple vendors' DPUs and IPUs can be provisioned, managed, and accessed in a common and standardized way, eliminating vendor lock-in and compatibility issues. We will dive into the different working groups, their focus areas, and the progress they have made in developing APIs for storage, networking, security, AIML, and more. Additionally, we will discuss the simulation environment and developer platform provided by the project, enabling developers to test and deploy applications in the OPI ecosystem. Finally, we will provide information on how interested individuals and organizations can join the project and contribute to its development.

Overview of the Open Programmable Infrastructure Project (OPI)

The Open Programmable Infrastructure Project (OPI) is an industry-wide initiative aimed at creating an open ecosystem for data processing units (DPU) and intelligent processing units (IPU). The project is hosted by the Linux Foundation and brings together a diverse group of industry leaders, including Dell, F5, Intel, Keysight, Marvel, Nvidia, Red Hat, Tencent, and ZTE, among others. The project's goal is to develop standardized APIs, protocols, and best practices that enable the seamless integration of DPUs and IPUs from different vendors, ensuring interoperability, portability, and greater flexibility for end-users.

Working Group Leaders and Their Roles

The OPI project is organized into several working groups, each led by industry experts who focus on specific areas of development. The working group leaders play a crucial role in driving the project's goals and facilitating collaboration between member organizations. Here are the working group leaders and their roles:

1. Joseph White (Dell): Fellow with Dell and OPI TSC Chair

   • Responsible for overall project coordination and management
   • Works closely with other working group leaders to align goals and priorities

2. Paul Pindell (F5): Principal Architect and Business Development Lead

   • Oversees technical aspects of partner solutions
   • Works on open ecosystem development and standardization efforts

3. Boris Blumcher (Dell): Distinguished Engineer and Life Cycle Provisioning Working Group Leader

   • Focuses on the provisioning and lifecycle management of DPUs and IPUs
   • Leads the development of standardized APIs and processes for device provisioning and firmware updates

4. Stephen Royers (Red Hat): Principal Software Engineer and POC/Developer Platform Working Group Leader

   • Leads the development of proof-of-concept projects and the developer platform
   • Provides guidance on the simulation environment and testing frameworks

5. Mark Sanders (Dell): Distinguished Engineer and API Working Group Leader

   • Leads the development of standardized APIs for storage, networking, security, and AIML
   • Ensures compatibility and interoperability across different vendor implementations

6. Additional leaders not mentioned in the content include team leaders for the networking, security, storage, and AIML working groups. Each leader focuses on their respective area of expertise and drives the development of standardized protocols and interfaces.

About SNIA

The Storage Networking Industry Association (SNIA) is an organization that promotes the adoption of storage networking standards and technology. SNIA has been in existence for 25 years and has grown from focusing solely on storage connectivity to covering all aspects of server storage and networking. It comprises over 180 member organizations and has over 2,500 contributors who actively participate in the development of storage-related initiatives. SNIA's mission is to provide education and informational resources to its members and the broader industry, fostering the adoption of interoperable storage solutions.

Legal Disclaimer

The information shared in this article is intended for educational and informational purposes only. No warranties or guarantees are implied, and the information should not be construed as professional advice or recommendations. The OPI project and its participants assume no responsibility for any actions taken based on the information presented.

Provisioning of DPU and IPO Devices

The provisioning of DPUs and IPUs is a critical aspect of the Open Programmable Infrastructure Project (OPI). The project aims to establish standardized processes and protocols for provisioning these devices to ensure secure, automated, and vendor-agnostic deployments. The primary focus areas of the provisioning working group include:

1. Secure Zero-Touch Provisioning: OPI adopts RFC A572, which defines a secure and automated provisioning process for DPUs and IPUs. This process ensures that devices are validated by the network, authenticate the network, and secure the exchange of artifacts between the devices and the bootstrap server.

2. Inventory Collection: The working group is developing standardized methods for collecting inventory information from DPUs and IPUs. This includes local inventory accessed from within the devices, remote network inventory retrieval, and inventory retrieval by the host or BMC.

3. Boot Sequencing: To ensure proper boot coordination, the working group is developing mechanisms for host-DPU boot sequencing. This involves halting the host until the DPU completes booting and establishing communication between the DPU and the host to release the hold and continue booting.

4. Lifecycle Management: OPI addresses various aspects of lifecycle management, such as firmware, software, and application upgrades, rebooting, factory resets, and recovery to a known state. The working group is developing standardized APIs and processes for these lifecycle operations.

5. Monitoring and Telemetry: OPI adopts the OpenTelemetry standard to enable consistent and vendor-agnostic monitoring and telemetry across DPUs and IPUs. This ensures that metrics, traces, and logs from devices can be streamed to a centralized location for analysis and visualization using widely used tools like Prometheus and Grafana.

By standardizing provisioning processes and APIs, OPI aims to simplify device deployment, improve security, and enhance interoperability between different vendors' DPUs and IPUs.

Inventory Collection from DPU and IPO Devices

The inventory collection process is a crucial component of managing and maintaining DPUs and IPUs within the Open Programmable Infrastructure (OPI) environment. The goal of the inventory collection working group is to develop standardized methods and protocols for gathering inventory information from these devices. This includes fetching both local and remote inventory data for a comprehensive view of the device's capabilities and status. The working group focuses on three main areas:

1. Local Inventory: Local inventory refers to the collection of data regarding applications and services running directly on the DPUs and IPUs. The working group leverages standardization efforts like DMI SM BIOS to access system information, including hardware details and configuration. This enables users to query and retrieve relevant inventory information directly from the devices.

2. Remote Network Inventory: Remote network inventory focuses on enabling external applications to query and retrieve inventory data from DPUs and IPUs across a network. The working group explores standards like DMTF Redfish and gRPC-based inventory query protocols. These protocols allow for streamlined and vendor-agnostic retrieval of inventory information, facilitating easier management and monitoring of devices in large-scale deployments.

3. Host and SMBMC Inventory: The working group also addresses how the host or system management baseboard controller (SMBMC) can access inventory information from DPUs and IPUs. This includes defining standards for accessing device-specific data, such as system resources allocation and utilization. By providing a standardized approach, users can retrieve crucial inventory information from devices via the host or SMBMC, ensuring efficient management and resource allocation.

Through these inventory collection efforts, OPI aims to simplify the retrieval and utilization of inventory data from DPUs and IPUs, facilitating effective monitoring, troubleshooting, and optimization of device performance.

Boot Sequencing

The boot sequencing working group within the Open Programmable Infrastructure (OPI) project focuses on defining standardized processes and mechanisms for booting DPUs and IPUs in coordination with the host. Key areas of focus include:

1. Host-DPU Boot Coordination: As DPUs and IPUs become integrated into servers, it becomes necessary to coordinate the boot process between the host and these devices. The working group aims to establish mechanisms to ensure that the host halts booting until the DPUs and IPUs have completed their boot process. Once the devices are ready, communication is established from the DPU/IPU to the host to release the hold and continue booting. This coordination ensures that all components are up and running before the system proceeds with initialization.

2. DPU Reboot Notifications: In scenarios where the DPUs are rebooted, it is essential to provide notifications to the host and other components. This allows the host to prepare for any potential impact, such as resources being freed within the DPUs during reboot. The working group is developing mechanisms to enable efficient communication and notification exchange between the host and DPUs to ensure smooth reboot operations.

3. Host Reboot Notifications: Similarly, when the host initiates a reboot, it is important to notify the DPUs to release any allocated resources or take necessary actions before the host restarts. The working group is defining protocols and APIs to facilitate bidirectional communication between the host and DPUs to enable smooth host reboot processes.

4. Error Handling and Crash Recovery: Crashes and errors within DPUs and IPUs can lead to system unresponsiveness. The working group aims to address these scenarios by developing error handling mechanisms and crash recovery protocols. This includes mechanisms to gracefully handle DPUs panics, provide notifications to the host, and enable fast recovery and system stability.

By standardizing boot sequencing processes, OPI ensures proper synchronization and coordination between host and DPU/IPU booting activities, enhancing system reliability and performance.

Lifecycle Management

The lifecycle management working group within the Open Programmable Infrastructure (OPI) project focuses on defining standardized processes and APIs for managing the lifecycle of DPUs and IPUs. Key areas of focus include:

1. Rebooting: The working group aims to define standardized mechanisms for rebooting DPUs and IPUs. This includes different types of reboot operations, such as firmware reboot, OS reboot, and application restart. By establishing uniform procedures for rebooting, OPI simplifies the management and operation of these devices.

2. Updates and Upgrades: The working group is developing standardized APIs and processes for updating and upgrading various components within DPUs and IPUs. This includes firmware updates, OS updates, bootloader updates, and updates to DPU software or applications. By providing consistent and clear guidelines for updates and upgrades, OPI ensures compatibility and compatibility between different vendor implementations.

3. Recovery to a Known State: In certain scenarios, it may be necessary to restore DPUs and IPUs to a known state, similar to a factory reset operation. The working group is defining standardized procedures and methods for recovering DPUs and IPUs from unexpected states, ensuring easy restoration and alleviating potential issues caused by improper device configurations.

4. Debugging and Diagnostics: The working group is addressing the need for comprehensive debugging and diagnostic tools for DPUs and IPUs. This includes defining standardized APIs and protocols for collecting diagnostic information, generating logs, and facilitating efficient troubleshooting in case of issues or errors. By standardizing these processes, OPI enables easier debugging and diagnosis of device-related problems.

The lifecycle management efforts within OPI aim to streamline the management of DPUs and IPUs across their entire lifespan, ensuring smooth operations, efficient updates, and quick recovery from unexpected scenarios.

Monitoring and Telemetry

Monitoring and telemetry play a vital role in maintaining and optimizing the performance of DPUs and IPUs within the Open Programmable Infrastructure (OPI) ecosystem. The monitoring and telemetry working group focuses on defining standardized processes and interfaces for collecting, analyzing, and visualizing telemetry data from these devices. Key areas of focus include:

1. OpenTelemetry Adoption: OPI has adopted the OpenTelemetry standard for consistent and interoperable monitoring and telemetry. OpenTelemetry combines metrics and trace standards to provide unified observability across different systems and environments. The working group ensures that OPI adheres to the OpenTelemetry standard, facilitating seamless integration with existing monitoring and visualization tools.

2. Traces, Metrics, and Logs: The working group aims to enable the collection and streaming of traces, metrics, and logs from DPUs and IPUs to a centralized location. DPUs and IPUs generate valuable data related to system utilization, performance, and resource usage. By providing standardized interfaces and protocols for collecting and transmitting this data, OPI enables efficient monitoring, troubleshooting, and optimization of device performance.

3. OpenTelemetry Collector Integration: To facilitate the collection and aggregation of telemetry data, the working group has developed an example OpenTelemetry collector integrated with key components such as SPDK (Storage Performance Development Kit) and system monitoring. The OpenTelemetry collector acts as a gateway and aggregator, collecting data from various DPUs and IPUs and forwarding it to a centralized location for further analysis and visualization.

By standardizing monitoring and telemetry processes, OPI enables comprehensive monitoring of DPUs and IPUs, allowing for proactive performance management, rapid troubleshooting, and effective resource utilization.

Storage APIs

The storage APIs working group within the Open Programmable Infrastructure (OPI) project focuses on developing standardized APIs for storage-related operations in the DPU and IPU ecosystem. The aim is to provide seamless integration, configuration, and management of various storage devices and protocols. Key areas of focus include:

1. Front-End APIs: Front-end APIs enable interaction between the host and the emulated storage devices presented by DPUs and IPUs. These APIs define how storage devices, such as emulated NVMe devices, are configured and presented in terms of physical functions, virtual functions, and resource allocation. The working group is developing standardized front-end APIs to ensure consistent and vendor-agnostic device interaction.

2. Back-End APIs: Back-end APIs focus on the interaction between DPUs and IPUs and the actual storage infrastructure, such as storage arrays. These APIs define the parameters needed to connect to remote storage systems using protocols like NVMe over Fabric (NVMe-oF) over TCP or RDMA. The working group aims to develop standardized back-end APIs to streamline storage connectivity and enable interoperability.

3. Middle-End APIs: Middle-end APIs sit between the front-end and back-end and provide additional functionalities such as compression, encryption, and other value-added services. These APIs allow the DPU or IPU to apply middle-end services to the data stream before it reaches the back-end storage. The working group focuses on defining standardized middle-end APIs, which can be utilized for various storage-related enhancements.

The working group is implementing these APIs using different approaches. It leverages existing open-source projects like SPDK to develop grpc protobuf APIs translated to SPDK APIs. Additionally, vendor-specific bridges are created for Marvel and NVIDIA to ensure vendor compatibility and seamless integration.

By standardizing storage APIs, OPI enables efficient, secure, and vendor-agnostic storage management, allowing users to leverage different storage devices and protocols while maintaining interoperability.

Networking APIs

The networking APIs working group within the Open Programmable Infrastructure (OPI) project focuses on developing standardized APIs and protocols for networking operations in the DPU and IPU ecosystem. The aim is to facilitate seamless integration, configuration, and management of various networking functions and capabilities. Key areas of focus include:

1. Cloud-Facing APIs: Cloud providers often require standardized APIs to configure DPUs and IPUs within their infrastructure. The working group is developing cloud-facing APIs that allow cloud providers to interact with the devices, configure networking functions, and manage resources in a vendor-agnostic manner.

2. Telco APIs: Telco environments have specific networking requirements and protocols. The working group is working on developing standardized APIs tailored to Telco use cases, meeting the demands of service providers, and ensuring compatibility and interoperability across different Telco deployments.

3. Kubernetes Networking APIs: Kubernetes is a popular container orchestration platform. The working group focuses on providing APIs that enable seamless integration of DPUs and IPUs into Kubernetes networking environments. By developing Kubernetes networking interfaces, OPI envisions efficient and standardized deployment of networking services within a containerized environment.

The working group endeavors to support a variety of network service capabilities, including technologies like SONiC and VPP (Vector Packet Processing). It leverages existing API models such as OpenConfig and CNI implementations to ensure compatibility and consistency across different deployments.

Standardizing networking APIs allows for easy integration, configuration, and management of DPUs and IPUs within diverse networking environments, promoting interoperability and streamlined operations.

Security APIs

The security APIs working group within the Open Programmable Infrastructure (OPI) project focuses on developing standardized APIs and protocols for security-related operations in the DPU and IPU ecosystem. The aim is to provide an open and seamless integration of security functions, ensuring compatibility, efficiency, and flexibility. Key areas of focus include:

1. IPsec: The working group is developing APIs and protocols for IPsec, enabling the secure transmission of data over IP networks. IPsec is essential for securing network communications, and the standardized APIs ensure compatibility between different vendor implementations.

2. Rule-Based Filtering: Rule-based filtering involves defining policies and rules for filtering network traffic based on specific criteria. The working group is developing APIs to facilitate the configuration and management of rule-based filtering, allowing users to define and enforce their security policies across DPUs and IPUs.

The working group leverages existing standards and protocols to ensure compatibility and interoperability between devices from different vendors. APIs are developed, taking into account the unique requirements of DPUs and IPUs while aligning with established security practices.

By standardizing security APIs, OPI aims to enable easier integration and management of security functions within the DPU and IPU ecosystem, ensuring robust protection and secure operations.

AIML APIs

The AIML (Artificial Intelligence/Machine Learning) APIs working group within the Open Programmable Infrastructure (OPI) project is tasked with developing standardized APIs for AIML-related operations in the DPU and IPU ecosystem. While the group is still in its early stages, it aims to address the specific requirements and advancements in AIML technologies. Key areas of focus include:

1. AIML Functionality: AIML encompasses various applications and functionalities, including computer vision, machine learning models, and data analytics. The working group aims to define APIs that enable seamless integration and utilization of AIML functions within the DPU and IPU ecosystem.

2. Collaboration with ION: The working group is collaborating with the ION organization, which focuses on integrating DPUs, IPUs, and GPUs for AI-based applications. By combining efforts, both projects aim to develop APIs and frameworks that allow efficient distribution and coordination of workloads across these different processing units.

As AIML is a rapidly evolving domain, the working group encourages contributions and input from experts in the field to define APIs that address the unique requirements of AIML applications. Contributions are welcome from individuals, organizations, and research institutions to shape the future of AIML within the OPI ecosystem.

Simulation Environment and Developer Platform

The Open Programmable Infrastructure (OPI) project provides a simulation environment and a developer platform to enable developers to test and deploy applications within the OPI ecosystem. These tools offer flexibility and ease of access to test functionalities without relying on physical DPU or IPU hardware. Key features of the simulation environment and developer platform include:

1. Simulation Environment: The simulation environment allows developers to create test environments that mimic real DPU/IPU deployments. It leverages containerization technologies like Docker and Docker Compose to create isolated environments for testing and development purposes. The simulation environment encompasses all the major components of the OPI ecosystem, including APIs, protocols, provisioning, monitoring, and telemetry.

2. Developer Platform: The developer platform provides developers with the necessary tools, documentation, and resources to build applications that integrate with the OPI ecosystem. It offers a comprehensive development environment with support for various programming languages and frameworks. The platform facilitates efficient collaboration between developers by providing development guides, sample codes, and access to project resources.

3. Continuous Integration (CI): The simulation environment and developer platform are integrated into the OPI project's continuous integration (CI) process. As new features and changes are introduced, automated tests are run against the simulation environment to ensure proper functionality and compatibility. The CI process helps maintain the stability and quality of the OPI ecosystem.

The simulation environment and developer platform enable developers to validate and test their applications in a controlled environment, supporting rapid development, iterative improvements, and efficient collaboration within the OPI community.

How to Join the Project

Joining the Open Programmable Infrastructure (OPI) project is open to individuals and organizations interested in contributing to the development and advancement of standardized APIs and protocols for DPUs and IPUs. Here are a few ways you can get involved:

1. Formal Participation: The OPI project welcomes formal participation from organizations, which allows for deeper involvement in shaping project direction and contributing to technical decision-making. For formal participation, interested organizations can contact the Linux Foundation, which oversees the project, to explore membership options and benefits.

2. Informal Participation: Individuals and organizations can participate in the project's activities and discussions informally. This includes attending public meetings, joining mailing lists, and engaging with the project's working groups and communities. Most technical discussions, meetings, and resources are publicly available, ensuring transparency and open collaboration.

3. Contributions and Feedback: OPI encourages active contributions and feedback from the wider community. Individuals can participate by sharing their expertise, proposing new ideas, raising concerns, and providing feedback on project documentation, specifications, and code. Contributions can be made through various channels, including mailing lists, project repositories, and public meetings.

By joining the project and contributing to its development, participants can help shape the future of DPUs and IPUs, drive industry-wide standardization efforts, and promote interoperability and innovation within the Open Programmable Infrastructure.

Conclusion

The Open Programmable Infrastructure (OPI) project aims to create an open ecosystem for data processing units (DPU) and intelligent processing units (IPU), enabling seamless integration, management, and interoperability across different vendors' devices. The project's working groups focus on developing standardized APIs and protocols for storage, networking, security, AIML, and more. These efforts are driven by industry experts and aim to address critical use cases, provide a simulation environment and developer platform, and facilitate collaboration and contributions from individuals and organizations. By joining the project and contributing to its development, participants can help shape the future of DPU and IPU technologies, promote open standards, and drive innovation in the programmable infrastructure landscape.

Pros:

• The OPI project brings together a diverse group of industry leaders to drive the development of standardized APIs and protocols for DPUs and IPUs.
• The project aims to create an open ecosystem that promotes interoperability, portability, and flexibility.
• The working groups focus on critical areas such as provisioning, inventory collection, boot sequencing, lifecycle management, monitoring, and telemetry.
• The simulation environment and developer platform enable developers to test and deploy applications within the OPI ecosystem.
• The project encourages participation and contributions from individuals and organizations, fostering collaboration and innovation.

Cons:

• As an emerging project, there may be challenges in coordinating the efforts of different working groups and aligning priorities.
• The project's success depends on the active involvement and contribution of industry stakeholders, which may require ongoing engagement and coordination.

Overall, the OPI project has the potential to drive significant advancements in the programmable infrastructure space, enabling greater interoperability and flexibility for DPUs and IPUs. By providing standardized APIs and protocols, the project aims to address key use cases, simplify device provisioning and management, and foster an open and collaborative ecosystem.