Pub/sub channels are lightweight and efficient. You can have millions of unique channels without worrying about memory or CPU usage. Channels are automatically cleaned up when they are no longer used.

The socketcluster CLI tool exposes kubectl (Kubernetes) shortcut commands to make it easy to deploy your app to any Kubernetes cluster. All necessary Kubernetes .yaml files are provided.

After deploying your app to a Kubernetes cluster, you can scale indefinitely by using the kubectl scale deployment command to add more scc-worker and scc-broker instances as needed.

You can perform asynchronous operations anywhere along a socket's inbound or outbound stream without disrupting the message processing order. Streams behave like processing queues by default.

Awaiting for asynchronous actions along a socket's inbound or outbout streams can cause messages to pile up. This can be easily monitored on the back end using socket.getInboundBackpressure() and socket.getOutboundBackpressure().

SocketCluster supports JWT authentication by default. This type of authentication is well suited for WebSockets as it allows short-lived tokens to be renewed and pushed out frequently while allowing you to save on database lookups.

Middleware streams allow you to block socket connections using the MIDDLEWARE_HANDSHAKE middleware line. Individual socket actions can be blocked using the MIDDLEWARE_INBOUND and MIDDLEWARE_OUTBOUND lines.

Every data packet which is sent between a client and server can be delayed or transformed using MIDDLEWARE_INBOUND_RAW, MIDDLEWARE_INBOUND or MIDDLEWARE_OUTBOUND lines.

All data and events are consumed using asynchronous loops (e.g. for-await-of). Event listener callbacks are not required; this solves many problems related to code readability and maintainability.

Clients are optimized to handle lost connections seamlessly. For example, if a client loses the connection, channels attached to that socket will be put in a pending state and automatically resubscribe after the socket reconnects.

Without callbacks, asynchronous logic is easier to follow. Using async/await and for-await-of loops makes it more obvious which parts of the code are serial and which parts are parallel; it encourages a more declarative style of programming.

Sockets and channels do not need to be destroyed explicitly. They will be automatically marked for garbage collection as soon as they stop being used and are no longer referenced in the code.