NP batching/streaming service

[lukasolson]

Browsers limit the number of HTTP connections with the same domain name. Most modern browsers limit this to six connections. In practice, this means that only six parallel requests can occur simultaneously, with subsequent requests being queued up until one of the previous requests has responded.

This causes issues in Kibana where having more than six parallel requests is quite common (e.g. having a Dashboard with more than six panels that each make a request to Kibana).

This problem could be solved by introducing a batching service/endpoint, which would accept multiple "requests", then use chunked transfer encoding to stream the responses as they become available.

This is already something that is being done in Canvas (#32027). It would be nice if this were a service provided by core so that any application could take advantage of this and avoid hitting the connection limit.

Todo

• Create bfetch plugin bfetch #52888
• Implement response streaming bfetch #52888
• Implement request batching bfetch (2) #53711
• Use bfetch for expressions plugin server-side functions bfetch (2) #53711

Possible follow-up:

• Use bfetch for esaggs Expressions function
• Use bfetch in Search services.

Related

• [Discuss] Execute HTTP requests internally
• Move server-side interpreter code to Canvas #44326
• Expressions server-side #46906

---

Parent issue: #46909

[elasticmachine]

Pinging @elastic/kibana-platform

[elasticmachine]

Pinging @elastic/kibana-app-arch (Team:AppArch)