申请试用
HOT
登录
注册
 
Deep dive#5 Milvus 如何实现实时查询
0 点赞
0 收藏
2下载
Milvus.io
/
发布于
/
32
人观看

Deep dive是由Milvus社区发起的代码解析系列直播,针对开源数据库 Milvus 整体架构开放式解读,与社区交流与分享 Milvus 最核心的设计理念。

对本期内容感兴趣的小伙伴,想要和讲师实时QA,欢迎大家添加小助手微信:Zilliz-tech 备注”直播“加入讨论群与大家共同交流!

本期分享大纲:
1.Milvus 2.0 数据插入流程回顾
2.数据加载到querynode的流程
3.查询流程
4.QA

展开查看详情

1.Milvus Deep Dive #5 How milvus enables real-time queries

2.Speaker bio Xi Ge Software Engineer Milvus2.0 & Arctern contributer • Education: Master, Huazhong University of Science and Technology • Homepage: https://github.com/xige-16/xige-16.github.io • Interests: Reading/Running

3.Agenda • Data insertion and data storage process review • Load data into query node • Real-time data query • Q&A

4.Data insertion and data storage process review

5.Milvus Architecture Overview Separation of storage and calculation

6.Milvus insert and storage process ① ② ③ ④ ⑤

7.Milvus Reliability • Data in log broker (streaming) + Data in storage (batch) = Full Data • With the support of log brokers and storage, query nodes can query in real time

8.Load data into query nodes

9.Data management in Log broker

10.Data management in Object storage

11.Flowcharts of load collection

12.Historical data management on query node Index can only be loaded into the sealed segment Load balance Query node down

13.Incremental data management on query node Flow graph insert data into growing segment

14.Real-time data query

15.Query request l Service Timestamp: All data inserted before service ts can be queried, updated by tsafe l Travel Timestamp: Only messages inserted before (service ts – travel ts) are queried l Guarantee Timestamp: start doing query after service ts >= guarantee ts

16.Query execute process on query node

17.Global reduce on proxy

18.Thank You https://milvus.io https://github.com/milvus-io/milvus 扫码加入直播交流群 关注 Milvus 视频号 https://twitter.com/milvusio 与讲师实时QA 直播视频早知道

0 点赞
0 收藏
2下载