eKuiper brain storming

Jump to bottom Edit New page

ngjaying edited this page Jul 28, 2023 · 8 revisions

Design For Edge

Should always fit the usage in resource-constrained, bandwidth sensitive and mass deployment node environments.

Smaller footprint and higher throughput
** Daily performance benchmark** +3 +2 +3 +3
Self manageable: the management operation should be automatic in most cases
Sys table for configures, and aggregate metrics
Edge/cloud connection enhancement: cache, resume etc.
Measurable, monitor and alert
- UX: One click rule debug by providing mock data
- Customize metrics
Export diagnostic info by one command, like go env +2
Error handling
Rate limit

Multi-stream

Easier to connect streams
Merge messages

Optimize for schema-less data

Avoid performance reduction
Infer schema: more static validation (Rockset)

Serialization

Faster format
Partly decode: only decode part of the message for the required field
Encrypt/decrypt

SQL expressiveness

Define variable by let +1
Pattern Recognition
With CTE/subquery

Fault-tolerance

Source cache for back pressure +1 +1
Faster checkpoint

Cluster for LB/HA

Integration

Neuron: faster format + auto schema +3
EdgeX: faster format

Programable

SDK vs API
Modularize and module tool (xgo & xcaddy) +2 +2

Industry

No OTA: enhance plugin and AI model
AI by WASM edge
Predefine rule like sourcing

Connectors

grpc
websocket
http2 push: android 车机
ssh

Refactor

gorilla/mux -> chi
Runtime optimization: constant folding +1
Incremental window