[Download] Design and Implementation of TiFlash DeltaTree Index

Translated
1

Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: 【资料下载】TiFlash DeltaTree Index 的设计和实现

| username: YY-ha

How does TiFlash read data?

Today’s source code interpretation introduces the underlying working principles, shared by PingCAP R&D engineer Li Dezhu in “Design and Implementation of TiFlash DeltaTree Index”

Download Materials

Analysis of the Design and Implementation of TiFlash DeltaTree Index.pdf (1.2 MB)

Recording replay: TiFlash DeltaTree Index_哔哩哔哩_bilibili

The full review will be published soon, thank you for your patience!

Source Code Promoter Benefits

Rewards: Becoming a [Source Code Promoter] will earn you

  • 100 points, 100 experience points
  • A random draw for 2 [Source Code Promoters] to win a TiDB custom thermos cup

How to participate: Randomly share the link you are watching/replaying to your Moments, take a screenshot and reply to this post with the screenshot (don’t forget to include the image), and you will get a chance to win a random draw~

Thermos Cup Image

2
| username: ShawnYan | Original post link

First comment, thanks for Mr. Li’s sharing~~ :clap:

3
| username: Billmay表妹 | Original post link

So fast, hahahahaha

4
| username: YY-ha | Original post link

You’re really ahead of the curve!

5
| username: szza | Original post link

The red-marked areas in the PPT are not very precise.

It should use node.count[i] and node.child[i].

6
| username: szza | Original post link

I have a question,

Does the node[i].sid in the announcement refer to the sid at position 1 or the sid at position 2?

7
| username: YY-ha | Original post link

Got it, I have reported it to the internal teacher~

8
| username: lidezhu-PingCAP | Original post link

The main purpose of the pseudocode here is to clarify the process. I think both ways of writing are fine, as long as the meaning is clearly understood.

9
| username: lidezhu-PingCAP | Original post link

This is a classic tree traversal algorithm, starting from the root node, so the sid of any internal node may be used. Initially, the sid of the root node is used.