tikv/tikv

Support scanning multiple ranges

Open

#4,239 创建于 2019年2月20日

在 GitHub 查看
 (8 评论) (0 反应) (0 负责人)Rust (15,922 star) (2,189 fork)batch import
component/performancedifficulty/mediumhelp wantedsig/transactionstatus/mentortype/enhancement

描述

Feature Request

Scanning multiple ranges (in Coprocessor) is not efficient currently because for each range we need to rebuild the rocksdb iterator. This results in a strange case that scanning from a to z can be faster than scanning from a to c, from f to g, from x to z.

We can support scanning multiple (ordered) ranges in the mvcc scanner, by using the same rocksdb iterator. Each time for a new range, we can next() for several times and then seek() for it, just like what we did in current mvcc process.

贡献者指南

技术栈
rust
领域
databasebackendperformance
议题类型
performance
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
5
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
3-5 days
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
clear
前置要求
Rust programmingRocksDB basicsTikV internals
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
15
研究方向
The issue describes inefficiency in scanning multiple ranges due to rebuilding the RocksDB iterator for each range. To implement, study the current mvcc scanner implementation in the TikV codebase, specifically how it iterates over key value pairs. The proposed solution involves reusing the same RocksDB iterator by seeking to the start of each new range after a series of next() calls. Relevant files likely include the mvcc scanner module and the RocksDB iterator wrapper. Review existing comments in the issue for any proposed approaches or caveats.