joyent-portal/legacy/spikes/metrics-service

metrics-service

A service which produces random streams of data (firehose) of interpretable metric information for various subsystems that can or will be pulled out from a container. These are currently defined in RFD-0037.

This spike's purpose was to find the best pattern to expose such metrics service with mock data.

http1

• very simple API to push data from the server to the client
• plain HTTP

• requires a connection to each container - because it's not bidirectional, we can't subscribe to more containers on-demand

http2

• very simple API to push data from the server to the client
• multiplexing

• push doesn't really help us, because it's only useful for static resources

graphql

• common interface between all data resources
• filter received data

• api not finalised yet
• api not documented

ws (with nes)

• strong integration with Hapi
• allow to subscribe and unsubscribe on-demand on a single connection

• some networks might require fallback to pooling