Project 3: The Discovery Server (Scaling Beyond Multicast)

A 3-container ROS 2 system that uses a discovery server instead of multicast.

Quick Reference

Attribute Value
Difficulty Level 2: Intermediate
Time Estimate 1-2 weeks
Main Programming Language Bash / Python
Alternative Programming Languages XML
Coolness Level Level 2: Practical but Forgettable
Business Potential 4. The Open Core Infrastructure
Prerequisites DDS discovery basics, Linux networking, ROS 2 env vars
Key Topics DDS Discovery, Discovery Server Mode, Multicast & Network Segmentation

1. Learning Objectives

By completing this project, you will:

  1. Explain how DDS Discovery affects ROS 2 behavior in this project.
  2. Implement the core pipeline for Project 3 and validate it with a deterministic demo.
  3. Measure and document performance or correctness under at least one stress condition.
  4. Produce artifacts (configs, logs, scripts) that make the system reproducible.

2. All Theory Needed (Per-Concept Breakdown)

DDS Discovery

Fundamentals

DDS Discovery is the process of peers exchanging metadata to form the ROS graph without a master. In ROS 2, this concept defines how nodes coordinate, exchange data, and enforce guarantees. At a minimum you should be able to name the primary entities involved, identify where configuration lives, and explain how PDP and EDP influence behavior. When you debug a system, you will almost always inspect participant or endpoint first because those details surface mismatches early. The practical goal is to build a mental map that connects the API knobs you change to the wire-level or runtime effects you observe. If you can explain this concept without naming a single ROS 2 command, you know it as a systems principle rather than a tooling trick, which is exactly what you need for production robotics.

Deep Dive into the concept

A deeper look at DDS Discovery starts by tracing data from the API surface to the middleware. Every time you configure PDP or EDP, ROS 2 expresses that intent in the rmw layer, which then maps the intent into DDS-RTPS structures. The mapping is not always one-to-one: a single policy or field can affect multiple runtime behaviors, including buffering, matching, and timing. This is why a simple change in participant can cause a subscriber to stop receiving data, or why two vendors can discover each other but never exchange payloads. The useful diagnostic strategy is to observe the graph (who matched), then the transport (what packets appear), and finally the runtime state (queues, deadlines, timers).

Failure modes cluster around mismatched assumptions. If endpoint is configured incorrectly, you may see data on one machine but not another, or discover that messages arrive but are rejected silently. If discovery traffic is too restrictive, you will observe a graph that looks healthy but never transitions into active data flow. In embedded settings, this can appear as missed deadlines or watchdog resets rather than explicit errors. A robust design therefore includes explicit validation: log the effective policy, emit version identifiers, and test a known-good baseline before you change parameters. This project forces that discipline because you will create repeatable experiments and capture deterministic outputs, so you can explain not only what happened but why it happened.

How this fits on projects

This concept directly shapes how you implement and validate Project 3. You will configure it, observe it, and stress it under controlled conditions.

Definitions & key terms

  • PDP: PDP in the context of DDS Discovery and ROS 2 systems.
  • EDP: EDP in the context of DDS Discovery and ROS 2 systems.
  • participant: participant in the context of DDS Discovery and ROS 2 systems.
  • endpoint: endpoint in the context of DDS Discovery and ROS 2 systems.
  • discovery traffic: discovery traffic in the context of DDS Discovery and ROS 2 systems.

Mental model diagram (ASCII)

[User Code] -> [DDS Discovery] -> [rmw/DDS] -> [Wire/Runtime Effects]
       |             |               |                 |
   Config/API     Policies        Entities         Observability

How it works (step-by-step, with invariants and failure modes)

  1. A node configures the concept through API calls or config files.
  2. The rmw layer translates the settings into DDS/RTPS fields (PDP, EDP).
  3. Peers evaluate compatibility, matching, or timing using participant and endpoint.
  4. The runtime queues or state machines enforce the policy and emit data.
  5. Observability tools (logs, CLI, packet capture) confirm discovery traffic behavior.

Minimal concrete example

participant -> endpoint -> matched reader/writer

Common misconceptions

  • Assuming defaults are identical across vendors.
  • Believing that discovery implies data flow without validating compatibility.

Check-your-understanding questions

  1. Explain how DDS Discovery changes runtime behavior in ROS 2.
  2. Predict what happens if PDP conflicts with EDP.
  3. Why might two nodes discover each other but still exchange no data?

Check-your-understanding answers

  1. It alters matching, buffering, or timing constraints expressed via DDS/RTPS.
  2. The endpoints fail to match or drop messages due to incompatible policy/encoding.
  3. QoS or policy mismatch prevents writer-reader matching or delivery.

Real-world applications

  • network troubleshooting
  • discovery server setups

Where you’ll apply it

References

  • DDS discovery documentation
  • Cyclone DDS discovery notes

Key insights

  • DDS Discovery is the lever that connects configuration to observable system behavior.

Summary

This concept is the bridge between theory and runtime evidence. Mastery means you can predict outcomes, not just observe them.

Homework/Exercises to practice the concept

  1. Capture or log a minimal trace where this concept is visible.
  2. Change one policy/setting and predict the system impact before running it.
  3. Explain the failure mode you expect if the configuration is wrong.

Solutions to the homework/exercises

  1. The trace should show the concept-specific fields or events you expect.
  2. Your prediction should name which endpoints match and how latency/loss changes.
  3. A wrong configuration should lead to mismatch, dropped data, or timeouts.

Discovery Server Mode

Fundamentals

Discovery Server Mode is centralized discovery using servers to replace multicast when networks block it. In ROS 2, this concept defines how nodes coordinate, exchange data, and enforce guarantees. At a minimum you should be able to name the primary entities involved, identify where configuration lives, and explain how server locator and client locator influence behavior. When you debug a system, you will almost always inspect static discovery or initial peers first because those details surface mismatches early. The practical goal is to build a mental map that connects the API knobs you change to the wire-level or runtime effects you observe. If you can explain this concept without naming a single ROS 2 command, you know it as a systems principle rather than a tooling trick, which is exactly what you need for production robotics.

Deep Dive into the concept

A deeper look at Discovery Server Mode starts by tracing data from the API surface to the middleware. Every time you configure server locator or client locator, ROS 2 expresses that intent in the rmw layer, which then maps the intent into DDS-RTPS structures. The mapping is not always one-to-one: a single policy or field can affect multiple runtime behaviors, including buffering, matching, and timing. This is why a simple change in static discovery can cause a subscriber to stop receiving data, or why two vendors can discover each other but never exchange payloads. The useful diagnostic strategy is to observe the graph (who matched), then the transport (what packets appear), and finally the runtime state (queues, deadlines, timers).

Failure modes cluster around mismatched assumptions. If initial peers is configured incorrectly, you may see data on one machine but not another, or discover that messages arrive but are rejected silently. If configuration is too restrictive, you will observe a graph that looks healthy but never transitions into active data flow. In embedded settings, this can appear as missed deadlines or watchdog resets rather than explicit errors. A robust design therefore includes explicit validation: log the effective policy, emit version identifiers, and test a known-good baseline before you change parameters. This project forces that discipline because you will create repeatable experiments and capture deterministic outputs, so you can explain not only what happened but why it happened.

How this fits on projects

This concept directly shapes how you implement and validate Project 3. You will configure it, observe it, and stress it under controlled conditions.

Definitions & key terms

  • server locator: server locator in the context of Discovery Server Mode and ROS 2 systems.
  • client locator: client locator in the context of Discovery Server Mode and ROS 2 systems.
  • static discovery: static discovery in the context of Discovery Server Mode and ROS 2 systems.
  • initial peers: initial peers in the context of Discovery Server Mode and ROS 2 systems.
  • configuration: configuration in the context of Discovery Server Mode and ROS 2 systems.

Mental model diagram (ASCII)

[User Code] -> [Discovery Server Mode] -> [rmw/DDS] -> [Wire/Runtime Effects]
       |             |               |                 |
   Config/API     Policies        Entities         Observability

How it works (step-by-step, with invariants and failure modes)

  1. A node configures the concept through API calls or config files.
  2. The rmw layer translates the settings into DDS/RTPS fields (server locator, client locator).
  3. Peers evaluate compatibility, matching, or timing using static discovery and initial peers.
  4. The runtime queues or state machines enforce the policy and emit data.
  5. Observability tools (logs, CLI, packet capture) confirm configuration behavior.

Minimal concrete example

FASTDDS_DISCOVERY_SERVER=UDPv4:[192.168.1.10]:11811

Common misconceptions

  • Assuming defaults are identical across vendors.
  • Believing that discovery implies data flow without validating compatibility.

Check-your-understanding questions

  1. Explain how Discovery Server Mode changes runtime behavior in ROS 2.
  2. Predict what happens if server locator conflicts with client locator.
  3. Why might two nodes discover each other but still exchange no data?

Check-your-understanding answers

  1. It alters matching, buffering, or timing constraints expressed via DDS/RTPS.
  2. The endpoints fail to match or drop messages due to incompatible policy/encoding.
  3. QoS or policy mismatch prevents writer-reader matching or delivery.

Real-world applications

  • industrial networks
  • Wi-Fi networks with multicast blocked

Where you’ll apply it

References

  • Fast DDS discovery server guide
  • ROS 2 DDS tuning docs

Key insights

  • Discovery Server Mode is the lever that connects configuration to observable system behavior.

Summary

This concept is the bridge between theory and runtime evidence. Mastery means you can predict outcomes, not just observe them.

Homework/Exercises to practice the concept

  1. Capture or log a minimal trace where this concept is visible.
  2. Change one policy/setting and predict the system impact before running it.
  3. Explain the failure mode you expect if the configuration is wrong.

Solutions to the homework/exercises

  1. The trace should show the concept-specific fields or events you expect.
  2. Your prediction should name which endpoints match and how latency/loss changes.
  3. A wrong configuration should lead to mismatch, dropped data, or timeouts.

Multicast & Network Segmentation

Fundamentals

Multicast & Network Segmentation is how VLANs, subnets, and switches affect multicast visibility and ROS 2 discovery. In ROS 2, this concept defines how nodes coordinate, exchange data, and enforce guarantees. At a minimum you should be able to name the primary entities involved, identify where configuration lives, and explain how VLAN and IGMP snooping influence behavior. When you debug a system, you will almost always inspect router boundaries or TTL first because those details surface mismatches early. The practical goal is to build a mental map that connects the API knobs you change to the wire-level or runtime effects you observe. If you can explain this concept without naming a single ROS 2 command, you know it as a systems principle rather than a tooling trick, which is exactly what you need for production robotics.

Deep Dive into the concept

A deeper look at Multicast & Network Segmentation starts by tracing data from the API surface to the middleware. Every time you configure VLAN or IGMP snooping, ROS 2 expresses that intent in the rmw layer, which then maps the intent into DDS-RTPS structures. The mapping is not always one-to-one: a single policy or field can affect multiple runtime behaviors, including buffering, matching, and timing. This is why a simple change in router boundaries can cause a subscriber to stop receiving data, or why two vendors can discover each other but never exchange payloads. The useful diagnostic strategy is to observe the graph (who matched), then the transport (what packets appear), and finally the runtime state (queues, deadlines, timers).

Failure modes cluster around mismatched assumptions. If TTL is configured incorrectly, you may see data on one machine but not another, or discover that messages arrive but are rejected silently. If firewall is too restrictive, you will observe a graph that looks healthy but never transitions into active data flow. In embedded settings, this can appear as missed deadlines or watchdog resets rather than explicit errors. A robust design therefore includes explicit validation: log the effective policy, emit version identifiers, and test a known-good baseline before you change parameters. This project forces that discipline because you will create repeatable experiments and capture deterministic outputs, so you can explain not only what happened but why it happened.

How this fits on projects

This concept directly shapes how you implement and validate Project 3. You will configure it, observe it, and stress it under controlled conditions.

Definitions & key terms

  • VLAN: VLAN in the context of Multicast & Network Segmentation and ROS 2 systems.
  • IGMP snooping: IGMP snooping in the context of Multicast & Network Segmentation and ROS 2 systems.
  • router boundaries: router boundaries in the context of Multicast & Network Segmentation and ROS 2 systems.
  • TTL: TTL in the context of Multicast & Network Segmentation and ROS 2 systems.
  • firewall: firewall in the context of Multicast & Network Segmentation and ROS 2 systems.

Mental model diagram (ASCII)

[User Code] -> [Multicast & Network Segmentation] -> [rmw/DDS] -> [Wire/Runtime Effects]
       |             |               |                 |
   Config/API     Policies        Entities         Observability

How it works (step-by-step, with invariants and failure modes)

  1. A node configures the concept through API calls or config files.
  2. The rmw layer translates the settings into DDS/RTPS fields (VLAN, IGMP snooping).
  3. Peers evaluate compatibility, matching, or timing using router boundaries and TTL.
  4. The runtime queues or state machines enforce the policy and emit data.
  5. Observability tools (logs, CLI, packet capture) confirm firewall behavior.

Minimal concrete example

ip maddr show 239.255.0.1

Common misconceptions

  • Assuming defaults are identical across vendors.
  • Believing that discovery implies data flow without validating compatibility.

Check-your-understanding questions

  1. Explain how Multicast & Network Segmentation changes runtime behavior in ROS 2.
  2. Predict what happens if VLAN conflicts with IGMP snooping.
  3. Why might two nodes discover each other but still exchange no data?

Check-your-understanding answers

  1. It alters matching, buffering, or timing constraints expressed via DDS/RTPS.
  2. The endpoints fail to match or drop messages due to incompatible policy/encoding.
  3. QoS or policy mismatch prevents writer-reader matching or delivery.

Real-world applications

  • robot labs with segmented networks
  • multi-floor deployments

Where you’ll apply it

References

  • Cisco multicast docs
  • Linux multicast troubleshooting guides

Key insights

  • Multicast & Network Segmentation is the lever that connects configuration to observable system behavior.

Summary

This concept is the bridge between theory and runtime evidence. Mastery means you can predict outcomes, not just observe them.

Homework/Exercises to practice the concept

  1. Capture or log a minimal trace where this concept is visible.
  2. Change one policy/setting and predict the system impact before running it.
  3. Explain the failure mode you expect if the configuration is wrong.

Solutions to the homework/exercises

  1. The trace should show the concept-specific fields or events you expect.
  2. Your prediction should name which endpoints match and how latency/loss changes.
  3. A wrong configuration should lead to mismatch, dropped data, or timeouts.

3. Project Specification

3.1 What You Will Build

A 3-container ROS 2 system that uses a discovery server instead of multicast.

Included features:

  • Deterministic startup with explicit configuration.
  • Observability (logs/CLI output) that exposes discovery/data flow.
  • A reproducible demo and a failure case.

Excluded on purpose:

  • Full robot control stacks or SLAM pipelines.
  • Custom GUIs beyond CLI output.

3.2 Functional Requirements

  1. **Fast DDS CLI usage: **Fast DDS CLI usage -> Launching discovery server with fastdds.
  2. **Environment variables: **Environment variables -> Using ROS_DISCOVERY_SERVER.
  3. **Network isolation: **Network isolation -> Ensuring nodes cannot discover each other without the server.
  4. Deterministic startup: The project must start with a reproducible, logged configuration.
  5. Observability: Provide CLI or log output that confirms each major component is working.

3.3 Non-Functional Requirements

  • Performance: Must meet the throughput/latency targets documented in the benchmark.\n- Reliability: Must handle common network or runtime failures gracefully.\n- Usability: CLI flags and logs must make configuration and diagnosis obvious.

3.4 Example Usage / Output

$ fastdds discovery -i 0 -l 0.0.0.0 -p 11811
$ FASTDDS_DISCOVERY_SERVER=UDPv4:[192.168.1.10]:11811 ros2 run demo_nodes_cpp talker
$ ros2 node list
/talker

3.5 Data Formats / Schemas / Protocols

Fast DDS discovery server env
FASTDDS_DISCOVERY_SERVER=UDPv4:[192.168.1.10]:11811

3.6 Edge Cases

  • Server reachable but blocked port
  • Mixed multicast + server modes
  • Domain mismatch

3.7 Real World Outcome

By the end of this project you will have a reproducible system that produces the same observable signals every time you run it. You will be able to point to console output, captured packets, or bag files and explain exactly why the result is correct. You will also be able to force a failure and demonstrate a clean error path.

3.7.1 How to Run (Copy/Paste)

# Build
colcon build --packages-select project_3
# Run
source install/setup.bash
# Start the main node/tool
./run_project_3.sh

3.7.2 Golden Path Demo (Deterministic)

$ fastdds discovery -i 0 -l 0.0.0.0 -p 11811
$ FASTDDS_DISCOVERY_SERVER=UDPv4:[192.168.1.10]:11811 ros2 run demo_nodes_cpp talker
$ ros2 node list
/talker

3.7.3 Failure Demo (Deterministic)

$ FASTDDS_DISCOVERY_SERVER=UDPv4:[192.168.1.10]:11811 ros2 node list
[WARN] No nodes discovered (server unreachable)

4. Solution Architecture

4.1 High-Level Design

[Input/Config] -> [Core Engine] -> [ROS 2/DDS] -> [Observability Output]

4.2 Key Components

Component Responsibility Key Decisions
Discovery Server Central discovery endpoint with static locator Use vendor-supported server mode
Client Config Point ROS 2 nodes to server locator Disable multicast if needed
Verification Tools Validate discovery with ros2 CLI and packet capture Deterministic startup

4.3 Data Structures (No Full Code)

struct ServerLocator {
  std::string ip;
  uint16_t port;
};

4.4 Algorithm Overview

Key Algorithm: Core Pipeline

  1. Start discovery server
  2. Launch nodes with server locator
  3. Verify endpoint matching

Complexity Analysis:

  • Time: O(n) over messages/events processed
  • Space: O(1) to O(n) depending on buffering

5. Implementation Guide

5.1 Development Environment Setup

# Install ROS 2 and dependencies
sudo apt-get update
sudo apt-get install -y ros-$ROS_DISTRO-ros-base python3-colcon-common-extensions

5.2 Project Structure

project-root/
|-- src/
|   |-- main.cpp
|   |-- config.yaml
|   `-- utils.cpp
|-- scripts/
|   `-- run_project.sh
|-- tests/
|   `-- test_core.py
`-- README.md

5.3 The Core Question You’re Answering

“How can ROS 2 run in enterprise or cloud networks where multicast is blocked?”

5.4 Concepts You Must Understand First

Stop and research these before coding:

  1. DDS Discovery
    • What breaks if this is misconfigured?
    • How will you observe it?
  2. Discovery Server Mode
    • What breaks if this is misconfigured?
    • How will you observe it?
  3. Multicast & Network Segmentation
    • What breaks if this is misconfigured?
    • How will you observe it?

5.5 Questions to Guide Your Design

  1. How will you prove multicast is disabled?
  2. How will you isolate container networks?

5.6 Thinking Exercise

Sketch a network diagram showing which packets travel with and without the discovery server.

5.7 The Interview Questions They’ll Ask

  1. “What is the purpose of a DDS discovery server?”
  2. “How does ROS_DISCOVERY_SERVER work?”

5.8 Hints in Layers

Hint 1: Use the fastdds CLI

fastdds discovery --server-id 0

Hint 2: Set environment variables

export ROS_DISCOVERY_SERVER=127.0.0.1:11811

Hint 3: Confirm multicast is blocked

sudo tcpdump -i eth0 udp port 7400
# Expect no discovery packets when multicast is disabled

Hint 4: Use separate Docker networks Create isolated Docker networks to ensure nodes only connect via the discovery server.

5.9 Books That Will Help

Topic Book Chapter
Topic Book Chapter
Networking “TCP/IP Illustrated” Ch. 10
Systems “Linux System Programming” Ch. 2

5.10 Implementation Phases

Phase 1: Foundation (1-2 days)

Goals:

  • Reproduce the baseline example from the original project outline.
  • Validate toolchain, dependencies, and environment variables.

Tasks:

  1. Create the repository and baseline project structure.
  2. Run a minimal example to confirm discovery/data flow.

Checkpoint: You can reproduce the minimal example and collect logs.

Phase 2: Core Functionality (1-2 weeks)

Goals:

  • Implement the full feature set from the requirements.
  • Instrument key metrics and logs.

Tasks:

  1. Implement each component and integrate them.
  2. Add CLI/config flags for core parameters.

Checkpoint: Golden path demo succeeds with deterministic output.

Phase 3: Polish & Edge Cases (3-5 days)

Goals:

  • Handle failure scenarios and document them.
  • Create a short report/README describing results.

Tasks:

  1. Add error handling, timeouts, and validation.
  2. Capture failure demo output and metrics.

Checkpoint: Failure demo yields the expected errors and exit codes.

5.11 Key Implementation Decisions

Decision Options Recommendation Rationale
Transport UDP, shared memory, serial UDP for baseline Simplest to observe and debug
QoS Default, tuned Default then tune Establish baseline before optimization

6. Testing Strategy

6.1 Test Categories

Category Purpose Examples
Unit Tests Validate parsers and helpers Packet decoder, config parser
Integration Tests End-to-end ROS 2 flow Publisher -> Subscriber -> Metrics
Edge Case Tests Failures & mismatches Wrong domain ID, missing config

6.2 Critical Test Cases

  1. Test 1: Baseline message flow works end-to-end.
  2. Test 2: Configuration mismatch produces a clear, actionable error.
  3. Test 3: Performance/latency stays within documented bounds.

6.3 Test Data

Use a fixed dataset or fixed random seed to make metrics reproducible.

7. Common Pitfalls & Debugging

7.1 Frequent Mistakes

Pitfall Symptom Solution
QoS mismatch Discovery works but no data Align policies explicitly
Misconfigured env vars No nodes discovered Print and validate env on startup
Network filtering Intermittent data Check firewall and multicast settings

7.2 Debugging Strategies

  • Start from the graph: confirm discovery before tuning QoS.
  • Capture packets: validate that RTPS traffic appears on expected ports.

7.3 Performance Traps

If throughput is low, check for unnecessary serialization, small history depth, or lack of shared memory.

8. Extensions & Challenges

8.1 Beginner Extensions

  • Add verbose logging and a dry-run mode.
  • Add a simple configuration file parser.

8.2 Intermediate Extensions

  • Add metrics export to CSV or JSON.
  • Add automated regression tests.

8.3 Advanced Extensions

  • Implement cross-vendor compatibility validation.
  • Add chaos testing with randomized loss/latency patterns.

9. Real-World Connections

9.1 Industry Applications

  • Fleet robotics where reliability must be guaranteed under lossy Wi-Fi.
  • Industrial systems that require deterministic startup and clear failure modes.
  • ROS 2 core repositories (rcl, rmw, rosidl)
  • DDS vendors: Fast DDS, Cyclone DDS

9.3 Interview Relevance

  • Explain QoS compatibility and discovery failures.
  • Describe how to debug why nodes discover but do not communicate.

10. Resources

10.1 Essential Reading

  • “Mastering ROS 2 for Robotics Programming” (focus on the sections related to DDS Discovery)
  • ROS 2 official docs for the specific APIs used in this project

10.2 Video Resources

  • ROS 2 community talks on middleware and DDS
  • Vendor tutorials on discovery and QoS

10.3 Tools & Documentation

  • ROS 2 CLI and rclcpp/rclpy docs
  • Wireshark or tcpdump for network visibility
  • Project 2: Builds prerequisite concepts
  • Project 4: Extends the middleware layer

11. Self-Assessment Checklist

11.1 Understanding

  • I can explain DDS Discovery without notes
  • I can explain how QoS and discovery interact
  • I understand why the system fails when policies mismatch

11.2 Implementation

  • All functional requirements are met
  • Golden path demo succeeds
  • Failure demo produces expected errors

11.3 Growth

  • I can explain this project in a technical interview
  • I documented lessons learned and configs
  • I can reproduce the results on another machine

12. Submission / Completion Criteria

Minimum Viable Completion:

  • Golden path demo output matches documentation
  • At least one failure scenario is documented
  • Metrics or logs demonstrate correct behavior

Full Completion:

  • All minimum criteria plus:
  • Compatibility verified across at least two QoS settings
  • Results written to a short report

Excellence (Going Above & Beyond):

  • Automated regression tests for discovery/QoS behavior
  • Clear compatibility matrix or benchmark chart