1、SelectDB Cloud 与 Apache Doris 存算分离的思考与实践杨勇强飞轮科技 技术副总裁Apache Doris PMC 成员Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Su
2、mmit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023个人介绍SelectDB 技术副总裁,主要负责云原生版本产品研发曾就职于百度智能云存储部主任架构师(T9)具有 10 年云存储产品与架构经验本科毕业于南开大学 计算机科学与技术专业,硕士毕业于中国科学院计算技术研究所Doris Summit Asia 2023Do
3、ris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asi
4、a 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023目录1.存算一体 vs 存算分离2.Apache Doris 存算分离架构实现方案3.存算分离功能示例4.存算分离和存算一体适用场景Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Su
5、mmit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20231存算一体 vs 存算分离Doris Summit Asia 2023Doris Summit Asia 2023Doris Sum
6、mit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023D
7、oris Summit Asia 2023Doris Summit Asia 2023Apache Doris 系统架构 低成本:SSD-HDD 扩缩容:节点间数据副本自动均衡 负载隔离:Resource Group、Workload GroupDoris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023D
8、oris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算一体 vs 存算分离 部署简单:仅FE与BE进程,BE和FE都可以单独扩容 稳定可靠:不依赖共享存储系统 性能优异:计算节点访问本地存储 为什么还需要存算分离
9、?什么样的存算分离架构更能贴合用户需求?要克服那些挑战?Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Su
10、mmit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023计算和存储解绑,单独扩缩容计算资源波谷波峰,灵活弹性数据存储冷热效应明显低成本与资源弹性读写任务分离更彻底的业务隔离,解决不同业务间的相互影响以及资源抢占问题可靠性与负载隔离单一数据面向不同的分析负载使用数据快速移动、快速备份恢复数据共享云上基础设施逐步完善,提供可靠的共享存储完全按量付费,资源消耗更低云基础设施的成熟为什么需要存算分离Doris S
11、ummit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 202
12、3Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023从实际需求出发,2.0 版本降低成本提升弹性逐步迭代低成本Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris
13、Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20232.0 冷热数据分层,实现冷数据存储分离根据将冷热数据分别存储在成本不同的存储介质上,从原本的 SSD-HDD 增加到 SSD-HDD-OS 三层;云磁盘的价格通常是对象存
14、储的 5-10 倍,如果可以将 80%的冷数据保存到对象存储中,存储成本至少可降低 70%;通过冷数据 Compaction 实现数据的高效压缩,提供冷数据 Cache 加速冷热数据查询,降低成本的同时保持性能不受影响;Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit A
15、sia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20232.0 弹性计算节点,实现计算分离引入了无状态的计算节点 Compute Node,专门用于执行计算分析;不保存任何数据,集群扩缩容时无需进行数据分片的负载均衡,具有明显高峰的场景中可以灵
16、活扩容、快速加入集群分摊计算压力;用户数据往往存储远端存储中,执行查询时查询任务会优先调度到 Compute Node 执行,以避免内表与外表查询之间的计算资源抢占。Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2
17、023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20232Apache Doris 存算分离架构实现方案Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Sum
18、mit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023整
19、体改造思路负载隔离低成本数据共享存储成本大幅下降计算和存储可以独立弹性使用业务的波峰波谷调整计算资源读写分离业务隔离内部负载隔离元数据服务Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summi
20、t Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算分离整体架构Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia
21、 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023热数据本地 Cache,全量数据在对象存储成本最高降低90%存
22、算一体:全量数据*3*块存储价格存算分离:热数据*1*块存储价格+全量数据*对象存储价格最高可以节省90%以上灵活的Cache管理手动预热读写预热Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris S
23、ummit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023元数据层高可靠高可用高性能语义 API 更加模块和内聚meta service 无状态写入事务依赖 FDB,缩短了导入路径Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Dori
24、s Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia
25、2023Doris Summit Asia 2023数据导入1.数据进入协调者 BE2.数据分发到多个 BE3.数据写入 Cache4.数据写入 S35.读写分离 Cluster 预热 Cache只需处理单副本数据数据和BE没有固定的关系没有 publish,写入流程更短Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summ
26、it Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023数据查询多个 cluster 独立不命中 Cache 时从 S3 读数据命中时从本地 Cache 读数据弹性资源大幅降低成本Doris
27、Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20
28、23Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023性能Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summ
29、it Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20233存算分离功能示例Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit As
30、ia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算分离 多计
31、算集群实现工作负载隔离Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Dor
32、is Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算分离 计算的弹性扩缩容手动:可以选择 cpu 数据进行伸缩分时:可以指定一条扩容规则和一条缩容规则Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summ
33、it Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算分离 计算的弹性扩缩容 集群自动停机Doris Summit Asia 2023Dori
34、s Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia
35、2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算分离 性能Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Dor
36、is Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 20234存算分离和存算一体适用场景Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Dor
37、is Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023存算一体 对性能要求更高、对查询时
38、延更敏感 不需要极致弹性扩缩容 没有可靠的共享存储可用 仅简单体验或开发测试 业务线独立运行Doris,无专职DBA 存算分离 成本敏感型,追求性价比 已经在公有云上构建服务 拥有可靠的共享存储服务 要求极致弹性,需要运行在容器平台上 有专职的团队维护数据仓库平台两种架构的适用场景Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris
39、 Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Resource Group、Workload Group 与多计算集群的融合实现共享的高速缓存,与计算节点进一步分离时间旅行
40、统一 catalog后续规划Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023
41、Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023获取更多社区动态与最佳实践Doris Summit 峰会官网:doris- Doris Summit 峰会回放:https:/ Doris 官网:doris.apache.orgApache Doris GitHub: Doris 官方平台:Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Dor
42、is Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023Doris Summit Asia 2023