Q: Your engineering org is migrating production workloads from Intel x86_64 EC2 instances to AWS Graviton (ARM64) instances to reduce compute costs by 40%. In CI, developers use `docker buildx build --platform linux/amd64,linux/arm64` running on AMD64 GitHub runners via QEMU user emulation. The builds take 35 minutes per PR, frequently crashing with `qemu: uncaught target signal 11 (Segmentation fault)`. You must diagnose QEMU emulation bottlenecks and architect a distributed Buildx builder cluster with native ARM64 and AMD64 worker nodes.

QEMU user-mode emulation relies on the Linux kernel `binfmt_misc` module to translate every foreign CPU instruction on the fly. Emulating complex CPU instructions, multi-threaded memory barriers, and SIMD operations incurs an 8x to 20x performance penalty and triggers memory race bugs in compilers like GCC, Rust, and Go.

<!-- Build Emulation Path -->
AMD64 Host CI Runner
  └── binfmt_misc kernel hook
        └── QEMU user emulator (translates ARM64 instructions -> AMD64 instructions)
              └── Compiler / Package Manager (8x slower, susceptible to SIGSEGV!)

Create a multi-node BuildKit builder instance. Point Buildx to a native AMD64 node for `linux/amd64` builds and an SSH-connected native ARM64 Graviton node for `linux/arm64` builds, eliminating QEMU emulation entirely.

# Create builder instance
docker buildx create --name multi-arch-builder --use

# Append native ARM64 remote worker via SSH
docker buildx create \
  --name multi-arch-builder \
  --append \
  --node worker-arm64 \
  --platform linux/arm64 \
  ssh://ci-runner@arm64-builder.internal

# Verify builder cluster configuration
docker buildx inspect --bootstrap

Run the multi-platform build using Buildx. Buildx delegates each architecture build to its corresponding native hardware node, merges the outputs into an OCI image index (manifest list), and pushes the unified manifest to the registry.

# Build concurrently on native hardware
docker buildx build \
  --builder multi-arch-builder \
  --platform linux/amd64,linux/arm64 \
  -t ghcr.io/myorg/api-service:v2.0 \
  --push .

# Inspect generated OCI image index manifest
docker buildx imagetools inspect ghcr.io/myorg/api-service:v2.0

For languages with built-in cross-compilers (Go, Rust), run the build on the host architecture and cross-compile for target architectures without invoking QEMU or requiring foreign hardware.

# Dockerfile using native BuildKit cross-compilation
FROM --platform=$BUILDPLATFORM golang:1.22-alpine AS builder
ARG TARGETOS
ARG TARGETARCH
WORKDIR /src
COPY . .
# Compiles natively on host CPU targeting foreign arch instantly!
RUN CGO_ENABLED=0 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -o /bin/app .

FROM scratch
COPY --from=builder /bin/app /bin/app
ENTRYPOINT ["/bin/app"]