The document discusses the use of Rust in TiKV, a distributed key-value database created by PingCAP, highlighting its benefits such as memory safety and thread safety over C++ and Go. It covers TiKV's architecture, the timeline of its development, key technologies employed, and future plans for Rust advocacy in China. The document is presented by Tang Liu, chief architect of PingCAP, who emphasizes the importance of Rust for improved performance and stability in TiKV.

1. Rust in TiKV
tl@pingcap.com

2. About me
● Tang Liu
● Chief Architect of PingCAP
● Worked at Kingsoft and Tencent
● Open sources: LedisDB, go-mysql, etc.

3. Agenda
● Why do we use Rust in TiKV
● TiKV architecture introduction
● Key technology
● Future plan

4. What’s TiKV?
A distributed Key-Value database with:
● Geo-Replication
● Horizontal scalability
● Consistent distributed transaction
● Coprocessor for distributed computing
● Working with TiDB like Spanner with F1

5. We need a language with:
● Fast speed
● Memory safety
● Thread safety
● Binding C efficiency

6. Why not C++?
● Memory leak, dangling pointer, etc.
● Data race
● No official package manager
● Development cycle will be long...

7. Why not Go?
● GC
● Data race
● Goroutine runtime expenses
● CGO expenses

8. So we turned to Rust...

9. But Rust...
● Steep learning curve
● Long compiling time
● Lack of libraries and tools
● Few programmers know it in
China
Time
Rust
Learning Curve

10. Why Rust?
● Memory safety
● Thread safety
● No GC
● Calling C is fast
● Official package manager

11. We made a hard but great decision:
Use Rust!

12. TiKV Timeline
Start
2016-01-01
Open Source
2016-04-01
Production
2016-10
Beta4
2016-11
RC1
2016-12
GA
2017-11
RC3
2017-04
RC2
2017-02

13. TiKV Architecture
KV API Coprocessor
Transaction
MVCC
Raft KV
RocksDB

14. Multi-Raft
Region 1
Region 3
Region 5
Region 4
Region 3
Region 5
Region 2Region 1
Region 3
Region 2
Region 4
Region 1
Region 5
Region 2
Region 4
TiKV node 1 TiKV node 2 TiKV node 3 TiKV node 4
Raft
Group

15. Region 1
Region 3
Region 1
Region 2
Scale-out (initial state)
Region 1*
Region 2 Region 2
Region 3Region 3
Node A
Node B
Node C
Node D

16. Region 1
Region 3
Region 1^
Region 2
Region 1*
Region 2 Region 2
Region 3Region 3
Node A
Node B
Node E
1) Transfer leadership of “region 1” from Node A to Node
B
Node C
Node D
Scale-out (add new node)

17. Region 1
Region 3
Region 1*
Region 2
Region 2 Region 2
Region 3
Region 1
Region 3
Node A
Node B
2) Add Replica on Node E
Node C
Node D
Node E
Region 1
Scale-out (balancing)

18. Region 1
Region 3
Region 1*
Region 2
Region 2 Region 2
Region 3
Region 1
Region 3
Node A
Node B
3) Remove Replica from Node A
Node C
Node D
Node E
Scale-out (balancing)

19. Client
TiKV
KV Thread
Raft Thread
RocksDB
Request
Request
Request
Log TiKV Node
Log TiKV Node
Log TiKV Node
Data
Parse the
protocol
Apply the log
A simple
write flow
Executes the
transaction logic

20. Key technology
● Networking
○ Protocol Buffers (protobuf)
○ mio
○ gRPC
● Asynchronous framework
○ mio + callback
○ tokio-core + futures
○ thread pool + futures - cpupool

21. Key technology
● Storage
○ rust-rocksdb
● Monitoring tool
○ rust-prometheus
● Profile tool
○ jemallocator with prof feature
● Lint tool
○ clippy

22. TODO...
● Make TiKV more fast, like removing Box
● Make TiKV more stable, like introducing Rust sanitizer
● Contribute more Rust open source modules, like raft library, open-tracing, etc.
● Participate in other Rust projects more deeply, like rust-gRPC
● Write more articles about Rust on Chinese social media and organize more
Rust meetups
● Be a strong advocate of Rust in China

23. Thank you!