Add PD scheduling best practices and glossary

[anotherrachel]

What is changed?

This PR:

• adds introduction.md of best practices
• adds pd-scheduling.md to the best practices series
• adds glossary.md

Any related PRs or issues?

Which version does your change affect?

[Hoverbear]

Maybe these can be split up

[anotherrachel]

You mean they can be introduced separately?

[Hoverbear]

I think they can be separate entries in this glossary?

[Hoverbear]

Seems reference is more appropriate

[dcalvin]

Actually I think these are some good concept material (reference, too) which users may want to notice easily when they are learning about TiKV.

[Hoverbear]

In this and some further examples you use concrete examples, but above you used [limit], can we make it consistent?

[anotherrachel]

Really appreciate your suggestions. I'll work on it.

[anotherrachel]

Hey Ana, here I think we can make it consistent by using scheduler add/remove [scheduler name] but I am not sure with "operator [add | remove]". The examples are concrete, yes, but more accessible to users since this operation usually involves a specific region (or store). What's your opinion?

[Hoverbear]

I feel like most of the first half of this article should probably be something for learning as part of concepts maybe? The rest of it seems to be troubleshooting/issue resolving. It's good and useful stuff. :)

But... I'm not really sure where the "Best Practices" part of this article this. It doesn't go into recommending for example, how to improve a PD cluster from the default configuration to one optimized for your needs. It doesn't talk about how to decide which options/topologies are ideal for your needs. It doesn't talk about tradeoffs involved between these options, etc.

[dcalvin]

@lilin90 PTAL

I feel like most of the first half of this article should probably be something for learning as part of concepts maybe? The rest of it seems to be troubleshooting/issue resolving. It's good and useful stuff. :)

But... I'm not really sure where the "Best Practices" part of this article this. It doesn't go into recommending for example, how to improve a PD cluster from the default configuration to one optimized for your needs. It doesn't talk about how to decide which options/topologies are ideal for your needs. It doesn't talk about tradeoffs involved between these options, etc.

[lilin90]

I feel like most of the first half of this article should probably be something for learning as part of concepts maybe? The rest of it seems to be troubleshooting/issue resolving. It's good and useful stuff. :)

But... I'm not really sure where the "Best Practices" part of this article this. It doesn't go into recommending for example, how to improve a PD cluster from the default configuration to one optimized for your needs. It doesn't talk about how to decide which options/topologies are ideal for your needs. It doesn't talk about tradeoffs involved between these options, etc.

@Hoverbear Agree with you. Thanks! We'll consider giving some suggestions to those who'll write best practices.