The Nugget
- Slack efficiently delivers real-time messaging globally using a robust server architecture with components like Channel Servers, Gateway Servers, and Admin Servers, ensuring user connectivity and message delivery with minimal latency.
Make it stick
- 🌍 Slack sends millions of messages daily across various channels in real-time.
- 💡 Each user has a persistent websocket connection for receiving instant updates.
- ⚙️ Central to Slack's architecture are Channel Servers (CS) and Gateway Servers (GS) which manage message routing and user subscriptions.
- ⏱️ Message delivery is typically completed in 500ms, showcasing the system's speed and efficiency.
Key insights
Real-time Messaging Architecture
Channel Servers (CS):
- Statefully manage channel histories.
- Host up to 16 million channels per server.
- Utilize consistent hashing for mapping channels, allowing quick recovery from server issues (under 20 seconds).
Gateway Servers (GS):
- Statefully manage user sessions and channel subscriptions.
- Deployed across multiple geographic regions for optimal user connection.
Admin Servers (AS):
- Stateless servers that interface between Slack's backend and channel servers.
Presence Servers (PS):
- Track online user statuses, enabling real-time presence notifications.
Message and Event Handling
Message Flow:
- Clients send messages via websocket connections to Gateway Servers, which then route to the appropriate Channel Server.
- The message is broadcasted to all clients subscribed to the channel in real time.
Event Processing:
- Different types of events trigger real-time updates (e.g., reactions, member additions).
- Spikes in event traffic are often linked to scheduled reminders and messages sent at the top of the hour.
Scalability and Future Growth
- Slack's architecture currently supports tens of millions of channels and clients, with room for further scalability to accommodate larger customer bases.
- The architecture is designed to handle future growth while maintaining performance and reliability.
Key quotes
- "If we look at the traffic on a typical work day, it shows that most users are online between 9am and 5pm local time."
- "Our message stats shows that the multiplicative factor for message broadcast is different across regions."
- "Slack serves tens of millions of channels per host, tens of millions of connected clients, and delivers messages globally in 500ms."
- "A single Slack team has all of its channels mapped across all the CSs."
- "With the linear scalability of our current architecture, we can serve many more customers."
This summary contains AI-generated information and may have important inaccuracies or omissions.