2025-07-24 Linaro Data Center Group steering committee Meeting Agenda/Minutes

Mathieu

Atsushi Ninomiya

Iakov

Randy

Joyce

When? 4th Thursday of each month.

Zoom connection details

• Meeting ID: 938 5333 7073

• Computer: https://linaro-org.zoom.us/j/93853337073

• Dial by your location

  • Find your local number: https://linaro-org.zoom.us/u/adhLE3WFEg

• Previous meeting minutes: INSERT

• Arm CCA enablement on OpenStack

• Project Updates

• Events

• AOB

• INSERT

• Arm CCA enablement on OpenStack (presented by Atsushi Ninomiya, from Fujitsu)

• Project Updates

  • Deploy CCA on Arm Platforms (DCAP)

    • Updates since last meeting:

      • Memory Encryption Contexts (MEC)

        • Patches with initial support submitted to QEMU devel mailing list.

      • Initial patches for very basic Device Assignment (DA) support shared with documention.

      • Standardization

        • Many ongoing discussions happening this week at the IETF 123 meeting.

    • Next tasks:

      • Add support for RME 1.1 new features in the CCA software stack:

        • Device assignment.

        • Memory encryption.

        • Planes

      • Kata containers

        • Transition Arm64 CI from Works on Arm Equinix environment to Linaro’s environment.

        • Produce proper documentation and scripts to run Kata containers with RME support.

        • Set up Kata Arm64 runtime-rs testing job.

      • Standards

        • Continue the work on CCA Token standarisation.

        • Continue the work on Endorsement API proposal.

        • Continue the work on the CCA Token extension for device assignment.

  • Performance Optimization (PO)

    • Updates since last meeting:

      • Worked on POLY1305: analysed the Neon implementation (and the corresponding paper by Bernstein on which it was based), and figured how to port it to SVE2. SVE(1) is unfortunately not an option as it doesn't support the 32bit*32bit -> 64bit widening multiplications and additions, which constitute the core of the efficient implementation. It is possible with SVE2. Unfortunately, it seems that SVE2 will have an unavoidable overhead if compared to Neon due to the fact that some instructions have higher latency than their Neon counterparts, and some Neon instructions have to be replaced by 2 or more SVE2 instructions, so there are simply more instructions in the resulting implementation. Finished porting, clened it up, built it, and now debugging, using OpenSSL's internal tests.

    • Next tasks:

      • Continue on POLY1305 porting.

      • Select a next optimization target (inside or outside OpenSSL).

  • Big Data & Data Science (BDDS)

    • Updates since last meeting:

      • Investigated Apache Doris on openEuler and its SIMD support.

    • Next tasks:

      • Apache Gluten Arm64 CI.

      • Apache Doris Arm64 CI.

  • Enhanced Software-Defined-Storage on Arm Servers (STOR)

    • Updates since last meeting:

      • Add minor fixes for Arm64 on SPDK.

    • Next tasks:

      • Debugging issues with Go affecting storage frameworks.

• Events

• AOB