DPDK Summit 2025

Containerized router is a software router accelerated for giving high performance using the DPDK library. A fundamental requirement for any router (software or hardware) is Quality of Service. In the case of hardware routers, the QOS is implemented in hardware ASICs. Many off-the-self NICs also support it, but it is highly dependent on the capability of the NICs. Moreover, in cloud environments, software NICs won’t have any such capability. So, for cloud native routers, it is best to implement it in software using software queues.

In this talk, we will describe how we are leveraging the DPDK APIs to provide end-to-end quality of service including classification, policing, rewrite, scheduling and shaping for providing QOS in a cloud native software router environment. We will also describe a new and scalable packet processing model for QOS packet scheduling. The end result is a highly scalable, optimal and flexible model to achieve software QOS for use-cases like 5G front haul and mid haul.

This session presents the Data Center Smart Switch use-case and looks at how the SONiC DASH pipeline can be implemented with a combination of flexible HW pipeline and programmable CPU cores on a smart NIC device, with the Intel IPU E2100 adapter provided as example.

Here DPDK can play a leading role in efficiently implementing some of the special stages of the SONiC DASH logical pipeline such as the connection tracking block that have critical performance requirements (as measured in millions of TCP/UDP connections per second) and this session summarizes the key learnings for a performant DPDK-based connection tracking implementation.

In this presentation, we explore the integration of advanced RISC-V extensions to enhance the performance and efficiency of DPDK.

1. CRC Implementation with Zbc Extension:
The RISC-V Zbc extension introduces instructions for carry-less multiplication, which can be leveraged to implement Cyclic Redundancy Check (CRC) in hardware.

2. Efficient Waiting with Zawrs Extension:
We demonstrate how to use Zawrs instructions to implement RISC-V-specific versions of the rte_wait_until_equal_* functions, including handling 16-bit values through pointer rounding and bit shifting. The Zawrs extension is also applicable to the DPDK power management library.

3. Prefetching with Zicbop Extension:
The RISC-V Zicbop extension introduces instructions for cache block operations, which can be utilized to implement the rte_prefetch* family of functions.

Through detailed examples and performance benchmarks, we illustrate the benefits of these RISC-V extensions in optimising DPDK operations. Attendees will gain insights into the practical implementation of these techniques and their impact on performance and energy efficiency in data plane applications.

There have been changes to ARM build system to ensure users get the full performance possible from their build platforms. At the same time, care has been taken to ensure good out of box experience. This talk will discuss ARM ISA revisions, how to create binaries for portability and for performance. It will also discuss typical issues users and developers might face and how to resolve them.

Feature arc abstraction[1] primarily aims to solve two use cases for rte_graph based applications: (a) It provides mechanism to add new protocols, as DPDK in-built nodes, which required to be enabled on a per interface basis (Example: IPsec etc.). Here interface refers to both physical port and virtual interface like VLAN/VXLAN etc. (b) It provides mechanism to facilitate application reusing DPDK in-built nodes without any code duplication. It proposes to add hook points in certain in-built nodes where application can eject packets (at a given networking layer) and may again inject same packet to different networking layer, after processing it.

This session will introduce usability and applicability of feature arc in rte_graph nodes and will try to discuss design decisions taken to achieve the same.

[1]: https://mails.dpdk.org/archives/dev/2025-January/311166.html

I have presented the project in Montreal last September and it received positive reactions from the DPDK community.

Since then, the project has received significant improvements and is now used internally at Red Hat for validating Telco infrastructure (for both OpenStack and OpenShift).

I would like to give a heads up to anybody who couldn't make it to Montreal. Also, I would like to challenge developers of core libraries about changes that would be beneficial to grout and DPDK.

The argparse library provides argument parsing functionality, it was integrated in DPDK 24.03, and makes it easy to write user-friendly command-line program. This topic will introduce this library usage.

The DPDK test suite (DTS) is a testing framework and set of end to end tests for DPDK which can be employed by DPDK developers and by labs performing CI testing. Given a minimal user configuration of the DPDK options and nodes to test, DTS will control the nodes, and test different DPDK workloads according to the test definitions. The framework will run a subset of the tests based on NIC capability checks completed at runtime, and provide a standard set of testrun logs and testrun result summaries.
In this talk, I will share how we have been leveraging DTS in the Community Lab. It is intended both to demonstrate a practical example of how the testing framework can and should be used, as well as provide an opportunity to discuss lessons learned over the past year and the development targets going forward for the DTS group. At DPDK summit 2024 there was a DTS presentation given which explained some of the DTS APIs exposed to testsuite writers, and explained how writing test suites worked. Ideally this talk will follow up on and compliment that talk by providing a real world example of DTS usage.

Requested time: 25 minutes (20 minute talk, 5 minutes for questions)

DPDK's collaborative development model continues to deliver an open and inclusive community, but it isn't without its challenges. Developers, eager to deliver new features as quickly as they can, will submit frequent versions of patches, including large series and large patches. A CI failure occuring in one patch leads to a quick respin and submit cycle. The result of this turnaround is an unintended effect: lots of wasted CI cycles, incomplete reviews, and a slower merging process.

In this talk, we'll explore the recent efforts to introduce intentional delays into the process to help upstream become more efficient. We will explain how small changes such as smarter individual testing, delayed submits, and improved communication can lead to a more sustainable and productive open-source development model.

If you've ever been frustrated waiting for the next CI result, or feel like the mailing list has become a bit 'write-only', this talk may be for you. Let's patch smarter, not harder!

As AI reshapes software development, DPDK can harness Retrieval-Augmented Generation (RAG) to provide developers with real-time access to best practices, optimization techniques, and expert knowledge. This session explores how AI-powered tools can help DPDK users improve performance tuning, troubleshoot issues, and generate optimized code by leveraging vast repositories of DPDK know-how. Additionally, we will demonstrate how AI can accelerate development by generating fast, tailored sample applications and test cases, enabling developers to prototype, validate, and optimize DPDK-based solutions more efficiently.