Nano Cluster is a set of tools to build, run and deploy Nano services (likes of function as a service, lambdas, serverless, ..etc.). Those Nano services are build and delivered as container images (possibly for WASM/WASI runtimes) Each Nano service only needs:

• STDIN: standard input to receive requests and results
• STDOUT: standard output to respond and send requests
• STDERR: to log

It uses Std.IO Universal Interface for message passing

The cluster orchestrator/dispatcher forks a process for the given container image typically the process will be isolated from host (not network access at all)

The process receives requests from STDIN and respond to STDOUT The most basic form of messages is a ndjson (a json in a single line) each message is a JSON-RPC. Other forms like msgPack, compact thrift, protoBuffer, flat-buffer will be added later.

By sending the request using STDOUT to special nano services called adaptors

{id:"123", method:"db.fetch_all", "params": {"sql": "select * from books"}}

• smaller footprint (a WASM service only needs fd_write and fd_read to be fully functional)
• fully Async IO
• fully event driven
• no network overhead
• no network saturation due to inter-container communications
• more secure (can't be sniffed, policy can be applied)

Check the FAQ