←All memories

Super-Sized Tech - A Look Inside McDonald's Event-Driven Architecture

 — #system-design#SDE

Hey everyone! When you think of McDonald's, your mind probably conjures up images of Big Macs, fries, and golden arches. But behind the fast-food empire lies a surprisingly sophisticated technology operation. To manage its vast global scale, McDonald's employs an impressive event-driven architecture to process a massive number of events efficiently.

It's a fascinating example of how even traditional giants are embracing modern tech principles. Let's take a peek at how they make it happen!

Beyond the Counter: The Need for Event Processing

With millions of customers, orders, and operational data points generated daily across the globe, McDonald's needs a robust system to handle this flow of information in real-time. An event-driven architecture (EDA) is a natural fit, allowing different parts of their system to react to occurrences (events) as they happen, leading to more responsive and scalable operations.

Standardizing the Flow: Key Components

To manage events effectively across their ecosystem, McDonald's has focused on standardization through several key components:

  • Event Registry: This is crucial for defining a standardized schema for all events. Consistency in event structure is vital for any large-scale event processing system.
  • Custom Software Development Kits (SDKs): McDonald's utilizes custom SDKs designed to process these standardized events and to handle any errors that may occur during processing.
  • Event Gateway: An event gateway serves as a central point for incoming events, performing essential functions like identity authentication and authorization before events are further processed.
  • Utilities and Tools: A suite of utilities and tools is in place for operational tasks such as fixing events, ensuring the health of the processing clusters, and performing various administrative functions.

Scaling for Global Operations: A Regional, Cloud-Powered Approach

Handling events from a global operation requires an architecture that can scale reliably. McDonald's achieves this with a regional architecture built on Amazon Web Services (AWS).

Here’s how they manage to scale their event processing:

  • Regional Deployment: The architecture is deployed regionally on AWS, which helps ensure global availability and resilience.
  • Domain-Based Event Sharding: Within each region, event producers (systems generating events) shard the events based on specific business domains. This means events related to, for example, orders might be handled differently or by different resources than events related to inventory.
  • MSK (Managed Streaming for Kafka) Clusters: Each of these sharded domains has its events processed by an AWS MSK cluster. MSK is a fully managed Apache Kafka service, which is well-suited for high-throughput event streaming.
  • Dynamic Auto-Scaling: To handle fluctuations in event volume, these MSK clusters are configured to auto-scale. This scaling is triggered based on various metrics, such as CPU usage of the Kafka brokers. The auto-scaling workflow itself is managed using AWS Step Functions and re-assignment tasks to adjust capacity as needed.

This approach allows McDonald's to process a high volume of events reliably, ensuring that different parts of its business can react to information in near real-time, from point-of-sale transactions to supply chain updates.

Why This Architecture?

While the source focuses on how McDonald's processes events, an event-driven architecture inherently offers several advantages for a business of this scale:

  • Decoupling: Services that produce events are decoupled from services that consume them, allowing them to evolve independently.
  • Scalability: As seen with their use of MSK and auto-scaling, EDA is well-suited to handling large and fluctuating event volumes.
  • Resilience: If a consuming service fails, events can often be queued and processed later, improving fault tolerance.
  • Real-time Responsiveness: Allows various systems to react to business occurrences as they happen.

Key Takeaways

  • McDonald's utilizes a sophisticated event-driven architecture to manage and process the vast number of events generated by its global operations.
  • They emphasize event standardization through an event registry, custom SDKs, and an event gateway for authentication/authorization.
  • The system is built on a regional AWS architecture, leveraging Managed Streaming for Kafka (MSK) for event processing.
  • Events are sharded by domain, and MSK clusters auto-scale based on operational metrics, managed by AWS Step Functions.

It's a powerful reminder that modern technology, including complex event-driven systems, underpins the operations of even the most familiar global brands.