Announcing the first open source Open Programmable Infrastructure (OPI) event

Lori MacVittie Miniatura
Lori MacVittie
Published March 10, 2022

With an infrastructure renaissance in full swing, the need for an open, standards-based DPU/IPU software stack is extant. The traditional building blocks of infrastructure have largely transitioned to software components to enable scale via general-purpose compute. The efficiency of this infrastructure, however, has slowly been decreasing as workloads grow more complex even as expectations of performance continue to expand. But the need to scale a wide variety of security, delivery, and network services and offset the slow but steady decline of Moore’s law is growing every day.

Because F5 believes it is important to foster an open ecosystem to collaborate in developing the next generation architectures and frameworks required to address this need, we drove the creation of the Open Programmable Infrastructure (OPI) project. The goal is to build a community-driven, standards-based DPU/IPU software stack capable of integrating with existing open source projects aligned with the same vision such as the Linux kernel, IPDK.ioDPDK, and SPDK.

Today, F5 is pleased to announce the first (virtual) OPI project event, which will be held on March 15–16, 2022 from 8 a.m. to 1:30 p.m. PT.

This is an exciting open (and free) opportunity to hear speakers from Intel, Google, Dell, Meta, F5, Cisco, VMware, IBM, Red Hat, Marvell, and others and learn how to join in the community effort underway to make infrastructure fully programmable across software and hardware devices such as IPUs, DPUs, and switches.

The full schedule, speaker list, and event links are available on the project site.  

What is the Open Programmable Infrastructure (OPI) Project?

DPU/IPU-like devices have their roots in the evolution of SmartNIC devices but separate themselves from that legacy in multiple important ways.

While a SmartNIC is clearly part of its host node’s compute system and exists to closely interact with and offload node hosted applications, the DPU/IPU dispenses with this secondary role. Instead, the DPU/IPU presents itself as a complete compute system unto itself with the following key characteristics:

  • Presence of their own general purpose processor
  • The ability to boot a general purpose OS
  • Domain-specific HW acceleration capabilities
  • Software-defined device functions that allow the software components deployed to them to define the device's functions that are presented to the host
  • Offloading of whole software subsystems, such as the Networking or Storage stack, including their control planes
  • Strict security isolation from the host on the hardware-level
  • Unique network identity
  • Out-of-band management where the Data/Infrastructure Processing Unit (DPU/IPU)-like device is managed separately from the server where it resides or the DPU/IPU-like device can be used to manage the server

But where will the right DPU/IPU software stack come from? And will it be a closed, walled garden, single vendor offering? Or will it be an open, collaborative, multi-vendor, innovation driven, ecosystem similar to what has occurred with the Kubernetes and Container environments?

The OPI project was created to address these questions and to foster the emergence of such an open and creative software ecosystem for DPU/IPU based cloud infrastructure.

You can learn more about the project here.