Add Docker MCP Toolkit Arm migration learning path

content/learning-paths/servers-and-cloud-computing/docker-mcp-toolkit/1-overview.md

@@ -0,0 +1,65 @@
+---
+title: The Arm migration challenge and how Docker MCP Toolkit solves it
+weight: 2
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Why migrate to Arm?
+
+Moving workloads from x86 to Arm64 has become a priority for organizations looking to reduce cloud costs and improve performance. AWS Graviton, Azure Cobalt, and Google Cloud Axion have made Arm-based computing mainstream, offering 20-40% cost savings and improved performance-per-watt for many workloads.
+
+The challenge is that migration is not always straightforward. For most containerized applications, rebuilding a Docker image for Arm64 is enough. But when you encounter legacy applications with hand-optimized x86 assembly, AVX2 intrinsics, or architecture-specific compiler flags, the migration becomes significantly more complex.
+
+## What makes x86-to-Arm migration hard?
+
+Traditional migration approaches require:
+
+- Manual code analysis for x86-specific dependencies.
+- Tedious compatibility checks across multiple tools.
+- Researching Arm NEON equivalents for each x86 intrinsic.
+- Rewriting vectorized code and adjusting loop structures.
+- Updating Dockerfiles, base images, and compiler flags.
+- Debugging compilation errors on the new architecture.
+
+For a single application with SIMD-optimized code, this can take 5-7 hours of manual work.
+
+## What the Docker MCP Toolkit provides
+
+The Docker MCP Toolkit is a management interface in Docker Desktop that lets you discover, configure, and run containerized MCP (Model Context Protocol) servers. It connects these servers to AI coding assistants through a unified gateway.
+
+For Arm migration, three MCP servers work together:
+
+- **Arm MCP Server**: Provides code scanning (`migrate_ease_scan`), Docker image architecture checking (`check_image`, `skopeo`), Arm knowledge base search (`knowledge_base_search`), and assembly performance analysis (`mca`).
+- **GitHub MCP Server**: Enables repository operations including creating pull requests, managing branches, and committing changes.
+- **Sequential Thinking MCP Server**: Helps the AI assistant break down complex migration decisions into logical steps.
+
+## How AI-assisted migration works
+
+When connected to the Docker MCP Toolkit, an AI coding assistant like GitHub Copilot can execute the entire migration workflow:
+
+1. Use `check_image` or `skopeo` to verify if base images support `linux/arm64`.
+2. Run `migrate_ease_scan` on the codebase to find x86-specific code, intrinsics, and compiler flags.
+3. Use `knowledge_base_search` to find correct Arm SIMD equivalents for every x86 intrinsic.
+4. Convert the code with architecture-specific accuracy.
+5. Update the Dockerfile with Arm-compatible base images and compiler flags.
+6. Create a pull request with all changes using the GitHub MCP Server.
+
+What normally takes 5-7 hours of manual work takes about 25-30 minutes.
+
+## What you will build in this Learning Path
+
+In this Learning Path, you will migrate a real-world legacy application - a matrix multiplication benchmark written with AVX2 intrinsics for x86 - to Arm64 using GitHub Copilot and Docker MCP Toolkit.
+
+The demo repository is available at [github.com/JoeStech/docker-blog-arm-migration](https://github.com/JoeStech/docker-blog-arm-migration).
+
+You will:
+
+1. Set up Docker MCP Toolkit with the Arm, GitHub, and Sequential Thinking MCP servers.
+2. Connect VS Code with GitHub Copilot to the MCP Gateway.
+3. Analyze the legacy x86 codebase to understand what blocks Arm migration.
+4. Use AI-driven tools to automate the full migration.
+5. Review the pull request created by the AI agent.
+
+In the next section, you will install and configure the Docker MCP Toolkit.

content/learning-paths/servers-and-cloud-computing/docker-mcp-toolkit/2-setup.md

@@ -0,0 +1,83 @@
+---
+title: Set up Docker MCP Toolkit with Arm, GitHub, and Sequential Thinking servers
+weight: 3
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Before you begin
+
+You need:
+
+- Docker Desktop 4.59 or later.
+- VS Code with the GitHub Copilot extension.
+- A GitHub account with a personal access token.
+- A machine with at least 8 GB RAM (16 GB recommended).
+
+## Enable the Docker MCP Toolkit
+
+1. Open Docker Desktop.
+2. Go to **Settings** > **Beta features**.
+3. Toggle **Enable Docker MCP Toolkit** on.
+4. Click **Apply**.
+
+After a few seconds, the **MCP Toolkit** tab appears in the left sidebar.
+
+## Add the required MCP servers
+
+Open the **MCP Toolkit** in Docker Desktop and select the **Catalog** tab. Add the following three servers:
+
+### Arm MCP Server
+
+Search for **Arm** in the catalog and add the [Arm MCP Server](https://hub.docker.com/mcp/server/arm-mcp/overview).
+
+Configure it by setting the directory path to your local code. This allows the `migrate_ease_scan` and `mca` tools to access your source files. Click **Save** after setting the path.
+
+The Arm MCP Server provides six tools:
+
+| Tool | Description |
+|------|-------------|
+| `knowledge_base_search` | Semantic search of Arm learning resources, intrinsics documentation, and software compatibility |
+| `migrate_ease_scan` | Code scanner for C++, Python, Go, JavaScript, and Java Arm compatibility analysis |
+| `check_image` | Docker image architecture verification for Arm64 support |
+| `skopeo` | Remote container image inspection without downloading |
+| `mca` | Machine Code Analyzer for assembly performance and IPC predictions |
+| `sysreport_instructions` | System architecture information gathering |
+
+### GitHub Official MCP Server
+
+Search for **GitHub Official** in the catalog and add the [GitHub MCP Server](https://hub.docker.com/mcp/server/github-official/overview).
+
+Configure authentication:
+
+1. Select the GitHub Official server.
+2. Choose **Personal Access Token** as the authentication method.
+3. Enter your GitHub token from **GitHub Settings** > **Developer Settings** > **Personal access tokens**.
+
+This server lets GitHub Copilot create pull requests, manage issues, and commit changes directly to your repositories.
+
+### Sequential Thinking MCP Server
+
+Search for **Sequential Thinking** in the catalog and add the [Sequential Thinking MCP Server](https://hub.docker.com/mcp/server/sequentialthinking/overview).
+
+No configuration is needed. This server helps GitHub Copilot break down complex migration decisions into logical steps.
+
+## Connect VS Code to the MCP Gateway
+
+1. In Docker Desktop, go to **MCP Toolkit** > **Clients** tab.
+2. Scroll to **Visual Studio Code** and click **Connect**.
+3. Open VS Code and click the **Extensions** icon in the left toolbar.
+4. Find **MCP_DOCKER**, click the gear icon, and click **Start Server**.
+
+## Verify the connection
+
+Open GitHub Copilot Chat in VS Code and ask:
+
+```text
+What Arm migration tools do you have access to?
+```
+
+You should see tools from all three servers listed. If they appear, your setup is complete.
+
+In the next section, you will examine the demo application to understand what blocks its migration to Arm64.

content/learning-paths/servers-and-cloud-computing/docker-mcp-toolkit/3-understand-the-demo.md

@@ -0,0 +1,86 @@
+---
+title: Understand the legacy x86 demo application
+weight: 4
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Clone the demo repository
+
+The demo application is a matrix multiplication benchmark written in C++ with AVX2 intrinsics. Clone it:
+
+```bash
+git clone https://github.com/JoeStech/docker-blog-arm-migration
+cd docker-blog-arm-migration
+```
+
+## Examine the Dockerfile
+
+Open the `Dockerfile`. There are two immediate blockers for Arm migration:
+
+**No Arm64 support in the base image**: The `centos:6` image was built for x86 only. This container will not start on Arm hardware.
+
+**x86-specific compiler flag**: The `-mavx2` flag tells the compiler to use AVX2 vector instructions, which do not exist on Arm processors.
+
+```dockerfile
+FROM centos:6
+
+RUN yum install -y \
+    devtoolset-2-gcc \
+    devtoolset-2-gcc-c++ \
+    devtoolset-2-binutils \
+    make \
+    && yum clean all
+
+WORKDIR /app
+COPY *.h *.cpp ./
+
+RUN scl enable devtoolset-2 "g++ -O2 -mavx2 -o benchmark \
+    main.cpp \
+    matrix_operations.cpp \
+    -std=c++11"
+
+CMD ["./benchmark"]
+```
+
+## Examine the source code
+
+Open `matrix_operations.cpp`. The code uses x86 AVX2 intrinsics throughout:
+
+```cpp
+#include <immintrin.h>  // x86-only header
+
+// Inside the multiply function:
+__m256d sum_vec = _mm256_setzero_pd();
+__m256d a_vec = _mm256_loadu_pd(&data[i][k]);
+sum_vec = _mm256_add_pd(sum_vec, _mm256_mul_pd(a_vec, b_vec));
+
+// Horizontal reduction
+__m128d sum_high = _mm256_extractf128_pd(sum_vec, 1);
+__m128d sum_low = _mm256_castpd256_pd128(sum_vec);
+```
+
+## Why this code cannot run on Arm
+
+There are several specific blockers:
+
+1. **x86-exclusive header**: `#include <immintrin.h>` only exists on x86 systems. Arm uses `<arm_neon.h>` instead.
+
+2. **AVX2 intrinsics throughout**: Every `_mm256_*` function is Intel-specific:
+   - `_mm256_setzero_pd()` creates a 256-bit zero vector (Arm NEON is 128-bit).
+   - `_mm256_loadu_pd()` loads 4 doubles at once (NEON loads 2 with `vld1q_f64`).
+   - `_mm256_add_pd()` and `_mm256_mul_pd()` are 256-bit operations (NEON uses 128-bit equivalents).
+   - `_mm256_extractf128_pd()` extracts the high 128 bits (not needed on NEON).
+
+3. **Vector width mismatch**: AVX2 processes 4 doubles per operation. Arm NEON processes 2 doubles per operation. The entire loop structure needs adjustment.
+
+{{% notice Note %}}
+SVE/SVE2 on newer Arm cores (Neoverse V1/V2, Graviton 3/4) provides 256-bit or wider vector-length agnostic (VLA) registers, matching or exceeding AVX2 width. The Arm MCP Server knowledge base can help determine the best approach for your target hardware.
+{{% /notice %}}
+
+4. **Horizontal reduction logic**: The pattern using `_mm256_extractf128_pd` and `_mm256_castpd256_pd128` is x86-specific and must be completely rewritten.
+
+Manual conversion requires rewriting 30+ lines of intrinsic code, adjusting loop strides, and testing numerical accuracy. This is exactly where the Docker MCP Toolkit with the Arm MCP Server becomes essential.
+
+In the next section, you will use GitHub Copilot with the Docker MCP Toolkit to automate the entire migration.

content/learning-paths/servers-and-cloud-computing/docker-mcp-toolkit/4-run-migration.md

@@ -0,0 +1,135 @@
+---
+title: Run the AI-driven Arm migration
+weight: 5
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Open the project in VS Code
+
+Open the cloned `docker-blog-arm-migration` directory in VS Code:
+
+```bash
+cd docker-blog-arm-migration
+code .
+```
+
+Make sure the MCP_DOCKER server is running in VS Code (check **Extensions** > **MCP_DOCKER** > **Start Server** if needed).
+
+## Give GitHub Copilot migration instructions
+
+Open GitHub Copilot Chat in VS Code and paste the following prompt:
+
+```text
+Your goal is to migrate this codebase from x86 to Arm64. Use the Arm MCP
+Server tools to help you with this migration.
+
+Steps to follow:
+1. Check all Dockerfiles - use check_image and/or skopeo tools to verify
+   Arm compatibility, changing the base image if necessary
+2. Scan the codebase - run migrate_ease_scan with the appropriate language
+   scanner and apply the suggested changes
+3. Use knowledge_base_search when you need Arm architecture guidance or
+   intrinsic equivalents
+4. Update compiler flags and dependencies for Arm64 compatibility
+5. Create a pull request with all changes using GitHub MCP Server
+
+Important notes:
+- Your current working directory is mapped to /workspace on the MCP server
+- NEON lane indices must be compile-time constants, not variables
+- If unsure about Arm equivalents, use knowledge_base_search to find docs
+- Be sure to find out from the user or system what the target machine is,
+  and use the appropriate intrinsics. For instance, if neoverse (Graviton,
+  Axion, Cobalt) is targeted, use the latest SME/SME2.
+
+After completing the migration:
+- Create a pull request with a detailed description of changes
+- Include performance predictions and cost savings in the PR description
+- List all tools used and validation steps needed
+```
+
+## Watch the migration execute
+
+GitHub Copilot orchestrates the migration through four phases using the Docker MCP Toolkit.
+
+### Phase 1: Image analysis
+
+Copilot uses the `skopeo` tool from the Arm MCP Server to analyze the base image:
+
+```text
+Checking centos:6 for arm64 support...
+```
+
+The tool reports that `centos:6` has no `linux/arm64` build available. This is the first blocker identified. Copilot determines that the base image must be replaced.
+
+### Phase 2: Code scanning
+
+Copilot runs the `migrate_ease_scan` tool with the C++ scanner:
+
+```text
+Running migrate_ease_scan with scanner: cpp
+```
+
+The scan detects:
+
+- AVX2 intrinsics (`_mm256_*` functions) in `matrix_operations.cpp`.
+- The `-mavx2` compiler flag in the Dockerfile.
+- The x86-specific header `<immintrin.h>`.
+
+Each issue includes the file location, line number, and specific code requiring modification.
+
+### Phase 3: Knowledge base lookup and code conversion
+
+For each x86 intrinsic found, Copilot queries the Arm MCP Server knowledge base:
+
+```text
+Searching knowledge base for: AVX2 to NEON intrinsic conversion
+```
+
+The knowledge base returns Arm documentation with the conversions:
+
+| x86 AVX2 Intrinsic | Arm NEON Equivalent |
+|---------------------|---------------------|
+| `_mm256_setzero_pd()` | Two `vdupq_n_f64(0.0)` operations |
+| `_mm256_loadu_pd()` | Two `vld1q_f64()` loads |
+| `_mm256_add_pd()` | Two `vaddq_f64()` operations |
+| `_mm256_mul_pd()` | Two `vmulq_f64()` operations |
+
+The knowledge base also explains that AVX2 uses 256-bit vectors processing 4 doubles at once, while NEON uses 128-bit vectors processing 2 doubles. Loop strides must be adjusted accordingly.
+
+Copilot rewrites the code using this information:
+
+- Replaces `<immintrin.h>` with `<arm_neon.h>` inside `#ifdef __aarch64__` guards.
+- Converts the AVX2 loop structure (stride 4) to NEON (stride 2).
+- Rewrites the horizontal reduction for NEON.
+- Updates the Dockerfile to use `ubuntu:22.04` with `TARGETARCH` for multi-arch builds.
+- Changes the compiler flag from `-mavx2` to `-march=armv8-a+simd`.
+
+### Phase 4: Create the pull request
+
+Copilot uses the GitHub MCP Server to create a pull request with:
+
+- All code changes (source files and Dockerfile).
+- A detailed description of what was changed and why.
+- Performance predictions for Arm.
+- A list of all MCP tools used during the migration.
+
+You can see an example PR at [github.com/JoeStech/docker-blog-arm-migration/pull/1](https://github.com/JoeStech/docker-blog-arm-migration/pull/1).
+
+## Summary of changes
+
+The migration produces these key changes:
+
+**Dockerfile**:
+- Replaced `centos:6` with `ubuntu:22.04`.
+- Added `TARGETARCH` for multi-architecture builds.
+- Changed `-mavx2` to `-march=armv8-a+simd` for Arm builds.
+
+**Source code**:
+- Added `#ifdef __aarch64__` architecture guards.
+- Replaced all `_mm256_*` AVX2 intrinsics with NEON equivalents (`vld1q_f64`, `vaddq_f64`, `vmulq_f64`).
+- Adjusted loop strides from 4 (AVX2) to 2 (NEON).
+- Rewrote horizontal reduction using NEON pair-wise addition.
+
+In the next section, you will build, test, and validate the migrated application on Arm.