unship commented 7 years ago

Summary

Add query plan graph to tidb's Web UI

Motivation

A sql query plan can be abstracted as a DAG which can be interactively visualized in browser as cockroach's init implementation of query plan graph @ QueryPlanGraph , and neo4j's implementation @ PlanView

Query Plan Graph can let user intuitively reasoning about how DB process sql to get data, and may help optimize their sql

Detailed design

Tensorflow is designed to be a computation engine which have some similarity with tidb stack, its monitor tool TensorBoard have DAG visualisation modul

Nodes and Edges can Represent operator and data flow

Symbol	Meaning
	High-level node representing a name scope. It can be a Raft group or tikv node, Double-click to expand to a set of operator running in this node
	Sequence of numbered nodes that are not connected to each other. It's a bunch of parallel operator
	Sequence of numbered nodes that are connected to each other. It's a bunch of sequential operator
	An individual operation node. It's a operator which have "next" method
	A constant. It's parameter of the tidb stack cluster
	A summary node. It's prometheus monitor value
	Edge showing the data flow between operations. It's grpc between nodes, it's width is throughput between nodes