SDKs & infrastructure
Horizontal scaling
Horizontal scaling adds more service instances to distribute work, rather than only increasing the resources assigned to one instance.
Also found under: Scale out, Load balancing
How it works
Persistent connections make routing and shared state important: a publisher may reach one server while its subscribers are connected elsewhere. Shared messaging infrastructure, connection accounting and failure handling must work across instances. A load balancer distributes connections but does not itself synchronize application events.
In a Pubb integration
Pubb separates HTTP API and Go socket services and uses Redis in its routing architecture. Self-hosted scaling still needs deployment-specific capacity testing; the architecture alone is not a guarantee of unlimited connections or throughput.
Explore plans and capacityA practical example
Two socket instances hold different clients. A shared routing layer lets a publication reach interested clients regardless of which instance accepted their connection.