
This is the final post in our Day in the Life of a Packet series on the Data Plane Development Kit (DPDK). Part 1 covered the technical foundations of DPDK’s architecture. Part 2 examined performance techniques including polling-mode drivers, zero-copy buffers, and NUMA-aware memory management.
In the previous parts of this series, we explored how the Data Plane Development Kit (DPDK) accelerates networking through polling-mode drivers, zero-copy buffers, NUMA-aware memory management, and related techniques. Behind these innovations lies a fifteen-year history of community collaboration that made DPDK viable as neutral, multi-vendor infrastructure.
This post looks at how DPDK evolved from an Intel-led initiative into a Linux Foundation project supported by a broad set of vendors, developers, and contributors, and how that transition shaped the toolkit now used across telecommunications, cloud, edge, scientific computing, and financial trading.
1. The Beginnings: Intel’s Vision for Network Optimization
DPDK’s journey began in 2010 when Intel introduced the framework as a toolkit to optimise data plane operations in high-performance network appliances. The initial focus was accelerating packet processing by bypassing traditional Linux networking stacks, reducing overhead, and enabling better performance in telecom and cloud environments.
Early work concentrated on Network Functions Virtualization (NFV), cloud-based networking, and high-throughput edge devices. The first versions of DPDK were tightly coupled with Intel’s hardware, particularly its Network Interface Cards (NICs).
At this stage, DPDK was largely an Intel-driven effort with contributions coming mainly from Intel engineers. The potential to change network performance economics became apparent quickly, and the broader community began to take notice.
2. The Transition to Open Source
Opening the Doors: Transition to the Linux Foundation
In 2013, DPDK transitioned from an Intel-only initiative to a fully open source project hosted at dpdk.org. This was a pivotal moment. DPDK became a neutral, vendor-agnostic project that invited contributions from a diverse set of companies, developers, and hardware vendors.
Another significant step forward occurred in 2016 when the community adopted a more formal governance structure. This led to DPDK becoming a Linux Foundation project in 2017, providing an open governance model and reinforcing the neutrality of the project.
Why Open Source Matters
The transition was about opening the doors to collaboration, innovation, and diverse contributions from the broader networking and cloud-native ecosystems. By adopting the open source model, DPDK became a unifying force for various industry players, encouraging multi-vendor support and balancing the influence of Intel and other vendors on the project’s direction.
The open source transition allowed DPDK to:
- Attract contributions from a wider community. Engineers from Red Hat, Arm, NVIDIA, Broadcom, Cisco, Marvell, NXP, and many others began contributing, bringing a range of insights, use cases, and optimisation strategies.
- Improve hardware support. With contributions from hardware vendors beyond Intel, DPDK gained the flexibility to support a broader range of NICs, processors, and accelerators. This enabled better performance across heterogeneous environments.
- Drive ecosystem growth. The ecosystem around the toolkit expanded rapidly, including DPDK-based solutions for telecom, enterprise, and cloud use cases, alongside user guides, training programs, and community events.
3. Community-Driven Innovation: Key Milestones
Milestone 1: Adoption Across Multiple Platforms
One of the reasons DPDK succeeded as an open source project was its ability to evolve across hardware platforms and software environments. By supporting a range of architectures including x86, Arm, and Power, and integration with Linux, FreeBSD, and Windows, DPDK became a cross-platform toolkit.
DPDK Growth Timeline
2010 to 2012 — Intel releases of DPDK under open source BSD license.
2013 — DPDK open source project founded at dpdk.org.
2014 — First contributions for non-Intel architectures, with Power8 support from IBM and ENIC support from Cisco. The first DPDK Summit was held in San Francisco.
2015 — The cryptodev API was added, expanding the scope of DPDK beyond Ethernet and marking the first step toward supporting a broader range of accelerators.
2016 — DPDK introduced Arm architecture support, opening new use cases in mobile and edge computing.
2017 — The DPDK project transitioned to the Linux Foundation, providing an open, neutral governance model.
2019 — The number of drivers in DPDK exceeded 100 for the first time, with 9 device types supported. Windows support was introduced upstream with the DPDK 19.05 release, adding a third operating system alongside Linux and FreeBSD.
2020 — DPDK 20.11 was released as a Long-Term Support (LTS) version, formalising the annual November LTS cadence and a three-year commitment to backported fixes for production deployments.
2022 — DPDK 22.11 LTS extended the November LTS cadence and consolidated significant flow API and mbuf modernisation work begun in the 21.11 cycle.
2024 — DPDK 24.11 LTS released, continuing the three-year LTS cadence with expanded support for high-speed NICs, SmartNICs, and accelerator devices.
2025 — DPDK marks 15 years since its 2010 launch. 103 new contributors participated in DPDK releases across the year (per LFX Insights), alongside continued expansion of libraries and drivers across networking, monitoring, and accelerator subsystems.
2026 — The 26.03 release, published in March, continues the project’s release cadence under Linux Foundation stewardship.
Milestone 2: Expanding Use Cases in Telecom and Cloud
DPDK was initially focused on telecom applications but quickly expanded to serve a variety of industries:
- Cloud and data centres. DPDK became a widely used solution for network acceleration in cloud platforms, powering applications requiring low-latency, high-throughput packet processing.
- Telecommunications. DPDK supported the NFV mission by enabling the disaggregation of software from hardware, allowing telecom service providers to run networking functions on industry-standard servers.
- Edge and 5G networks. DPDK provided the high-speed data plane needed for next-generation mobile and IoT services.
Milestone 3: Unlocking Unconventional and Emerging Use Cases
As DPDK matured and became more accessible through open source channels, developers outside traditional telecom and data centre domains began experimenting with its capabilities. The toolkit’s low-latency design and high-throughput characteristics made it a natural fit for use cases demanding real-time performance in areas beyond standard networking.
Radio Astronomy and Scientific Instrumentation. Research facilities like the South African Radio Astronomy Observatory (SARAO) have used DPDK to process radio telescope data, enabling real-time packet capture and filtering for petabytes of sky survey information. In these systems, DPDK supports custom UDP stream handling directly on FPGA-backed NICs without kernel overhead, which is critical for extremely high-throughput, low-latency workloads.
High-Frequency Financial Trading. Financial institutions rely on sub-microsecond latency to gain competitive advantages in market execution. Firms have adopted DPDK to create ultra-low-latency trading stacks that bypass the kernel entirely, leveraging direct NIC access and core pinning to process tens of millions of packets per second with deterministic performance. Solutions such as Solarflare/OpenOnload integrate with DPDK to enable this level of performance in latency-sensitive environments.
Satellite and Space Communications. Satellite ground stations have adopted DPDK for real-time beamforming, packet routing, and telemetry collection. DPDK’s ability to handle high-throughput UDP/IP streams at the edge, often in ruggedised, power-constrained environments, makes it well suited for satellite modems and RF-over-IP gateways. Companies in the “New Space” sector have also explored DPDK to optimise ground infrastructure for Low-Earth Orbit (LEO) satellite constellations.
Sensor Networks and Industrial IoT. In smart grid infrastructure and time-critical industrial automation, DPDK is used to process telemetry from distributed sensors. Applications such as real-time fault detection or predictive maintenance benefit from the deterministic packet processing DPDK offers, allowing systems to respond to anomalies within strict timing windows.
These use cases underscore how DPDK’s flexibility and performance have opened the door to new classes of applications that go far beyond its original telecom roots.
Milestone 4: Community and Ecosystem Growth
The DPDK community also grew because of its commitment to collaboration. Regular DPDK summits, community calls, and hackathons kept active discussions running on performance improvements, new features, and best practices. The open source model fostered an inclusive culture where ideas from all stakeholders could be implemented, leading to:
- Unified standards for network performance across vendors.
- Collaboration on testing frameworks to ensure cross-vendor interoperability.
- Industry-wide recognition as a primary toolkit for high-performance networking.
Milestone 5: Expanding the Data Plane Beyond Ethernet
DPDK’s original mandate was Ethernet packet processing. That scope has grown substantially. Successive releases have added support for cryptographic acceleration (the cryptodev API, introduced in 2015), compression (compressdev), baseband processing (bbdev), Direct Memory Access (DMA) operations, and event handling (eventdev). Each subsystem follows the architectural pattern that made the Ethernet device (ethdev) API successful: a stable API that abstracts underlying hardware differences, allowing applications to move across vendor implementations without a rewrite.
The practical result is that DPDK is now used well beyond its original network appliance context. Cryptographic offload supports TLS termination, IPsec, and PDCP. Baseband processing supports 5G radio access. DMA and event handling underpin latency-sensitive applications across finance, scientific instrumentation, and industrial automation.
“We constantly have new generations of products, on the CPU side, on the NIC side. Being able to rely on DPDK as the common abstraction layer that abstracts away those underlying differences is crucial.”
Tim O’Driscoll, Chair, DPDK Governing Board
The stability of the ethdev API and its sibling subsystems has become as central to DPDK’s adoption as the performance techniques it pioneered. Applications built on DPDK’s APIs move across CPU generations and NIC generations without rewriting.
4. The Power of Collaboration: Contributions from Industry Leaders
DPDK’s success as an open source project is also thanks to contributions from some of the largest hardware and software companies in the world. Key contributions came from:
- Red Hat. Integration with OpenStack, Kubernetes, and other cloud-native technologies.
- Arm. Support for Arm-based servers, expanding DPDK into mobile and embedded markets.
- Intel. Continued leadership and optimisation of DPDK for Intel architecture.
- NVIDIA. Integrating DPDK with GPU-accelerated networking to support data centres and AI workloads.
- Broadcom, Cisco, Chelsio, and others. Expanding DPDK to work with a range of NICs, from 10GbE to 100GbE and beyond.
The involvement of these vendors ensured that DPDK was optimised for a wide array of hardware platforms, making it a preferred choice for high-performance, scalable networking solutions across enterprise, telecom, scientific computing, financial trading, and aerospace systems.
5. A Look Ahead: DPDK’s Next Chapter
Fifteen years in, DPDK is positioned to keep evolving to meet the needs of next-generation networks. Several directions are already active work streams. Others sit as open technical questions that the community is still framing.
Near-term technical directions
- 5G and mobile core. DPDK’s role in 5G will continue to grow as telecom providers require ultra-low latency and high throughput for mobile networks and radio access.
- Edge computing. DPDK will continue to power edge devices where performance and efficiency are crucial for real-time data processing.
- Cloud-native networking. As cloud-native technologies like Kubernetes and service meshes continue to grow, DPDK’s integration with these platforms will remain a differentiator.
Open questions the community is working through
Alongside those near-term directions, a set of open technical questions is being discussed but not yet settled. Tim O’Driscoll, Chair of DPDK’s Governing Board, names several:
- AI infrastructure. Whether DPDK plays a role in AI networking infrastructure, and what that role looks like in practice, is under active discussion. The scale and consistency of data movement in AI workloads maps onto some of DPDK’s original strengths.
- The Ultra Ethernet Consortium (UEC). Formed in 2023 as a Linux Foundation Joint Development Foundation project, UEC published its 1.0 specification in June 2025 to redefine Ethernet for AI and High-Performance Computing (HPC) workloads. Whether DPDK becomes a component in a UEC software stack is one of the open conversations happening in and around the community.
- Cryptographic acceleration. Continued investment in DPDK’s crypto subsystem is aimed at expanding adoption in network security use cases, including higher-throughput TLS termination, IPsec offload, and preparation for post-quantum algorithms.
- Hyperscaler adoption. Broader adoption by hyperscale cloud providers is a direction of active interest, both as a use case that stresses the toolkit at scale and as a channel for upstream feedback and contribution.
- 6G. As 5G deployment matures, early work is beginning to identify where DPDK’s data-plane primitives fit into emerging 6G radio access and core network architectures.
The community priority
Alongside the technical roadmap sits a community question. DPDK’s contributor base skews heavily toward silicon and hardware vendors. End users, the companies building products on DPDK rather than selling the hardware it runs on, are under-represented on the Technical Board and in day-to-day mailing list discussions.
Tim frames it directly:
“It’s one of those things that’s easy to say and hard to do. But if we could get a broader base of DPDK users who are more active in the community, that would help to round things out.”
Tim O’Driscoll, Chair, DPDK Governing Board
Getting more end users, and more of the sectors that already use DPDK quietly (telecommunications, financial trading, scientific research, industrial automation), engaged in shaping the project is the through-line for the community’s next chapter.
Conclusion: Open Source and Community Innovation
DPDK’s growth from an Intel-specific project to a community-driven open source initiative has been transformative. By embracing the open source model with the guidance of the Linux Foundation, DPDK enabled a diverse set of stakeholders to contribute, resulting in better hardware support, cross-platform capabilities, and a broad range of use cases that go beyond what was originally imagined. The collaborative spirit of DPDK’s contributors, from vendors to independent developers, is what sets the project apart.
The next fifteen years hold work ahead for DPDK as it continues to address new demands in high-performance networking. This concludes the three-part blog series on DPDK, and we hope you have gained a deeper appreciation for how DPDK has become an important tool for networking in today’s cloud-native, virtualised world.
Get Involved
DPDK’s next chapter depends on the same thing the last fifteen years did: contributors and reviewers across the ecosystem. New reviewers are the highest-leverage way to help the project scale.
Start reviewing patches · Submit your first patch · Join a community call
Blog Contributors
Adrian Hoban, Tim O’Driscoll, Bruce Richardson, Georgii Tkachuk, Edwin Verplanke, Ben Thomas.

