数据同步机制自适应优化的HTAP数据库原型系统

doi:10.3969/j.issn.1000-5641.2023.05.002

摘要/Abstract

摘要：

在HTAP (hybrid transactional and analytical processing)数据库中, 资源隔离和数据共享是一个难题, 虽然不同厂商通过不同的架构来实现资源隔离, 但是用户关注的新鲜度, 即OLTP (online transaction processing)和OLAP (online analytical processing)读写版本的差距, 由数据共享的一致性模型决定. 然而, 现有的HTAP数据库为了节约成本, 只应用单一一致性同步模型, 这与用户应用的多种一致性需求之间存在矛盾, 为了满足用户需求的最高一致性而采取向上兼容的方案降低了系统的整体性能. 通过构建新鲜度与性能权衡的代价模型, 提出了一致性切换算法和切换前后同步数据的处理策略, 实现了一个顺序一致性同步与线性一致性同步自适应切换的HTAP数据库原型系统, 使得无需调整HTAP架构, 即可支持不同一致性(新鲜度)需求的查询负载并实现系统性能的最大化, 最后对自适应切换的有效性也进行了实验验证.

关键词: HTAP数据库系统, 数据同步, 一致性模型, 自适应同步模式切换, 性能与新鲜度权衡

Abstract:

In HTAP (hybrid transactional and analytical processing) database, resource isolation and data sharing is a difficult problem. Although different vendors achieve resource isolation through different architectures, the freshness of user concerns, that is, the gap between online transactional processing (OLTP) write and online analytical processing (OLAP) read versions, is determined by the consistency model of data sharing. However, existing HTAP databases apply only one consistency synchronization model for an easy implementation, which is contradictory to the multiple consistency requirements of user applications, and the overall system performance is sacrificed for the highest consistency upward compatibility. In this paper, by constructing a cost model of freshness and performance tradeoff, proposing a consistency switching algorithm and a processing strategy for synchronized data before and after switching, and realizing an HTAP database prototype with adaptive switching between sequential consistency synchronization and linear consistency synchronization, which makes it possible to support query loads with different consistency (freshness) requirements and maximize the system performance without adjusting the HTAP architecture. The effectiveness of adaptive switching is also verified by extensive experiments.s of adaptive switching is also verified by extensive experiments.

Key words: HTAP database systems, data synchronization, consistency model, adaptive synchronization mode switching, performance and freshness trade off

中图分类号:

TP392

俞融, 杨攀飞, 王清帅, 张蓉. 数据同步机制自适应优化的HTAP数据库原型系统[J]. 华东师范大学学报（自然科学版）, 2023, 2023(5): 11-25.

Rong YU, Panfei YANG, Qingshuai WANG, Rong ZHANG. An HTAP database prototype with an adaptive data synchronization[J]. Journal of East China Normal University(Natural Science), 2023, 2023(5): 11-25.

图/表 14

表1

图1

图2

图3

图4

图5

图6

图7

表2

表3

表4

图8

表5

图9

参考文献 21

1	LI G, ZHANG C. HTAP databases: What is new and what is next[C]// Proceedings of the 2022 International Conference on Management of Data. 2022: 2483-2488.
2	胡梓锐, 翁思扬, 王清帅, 等. HTAP数据库系统数据共享模型和优化策略 [J]. 软件学报, 2023. DOI: 10.13328/j.cnki.jos.006901.
3	LYU Z, ZHANG H H, XIONG G, et al. Greenplum: A hybrid database for transactional and analytical workloads [C]// Proceedings of the 2021 International Conference on Management of Data. 2021: 2530-2542.
4	HUANG D, LIU Q, CUI Q, et al.. TiDB: A Raft-based HTAP database. Proceedings of the VLDB Endowment, 2020, 13 (12): 3072- 3084.
5	WANG J Y, LI T L, SONG H Z, et al.. PolarDB-IMCI: A cloud-native HTAP database system at Alibaba. Proceedings of the ACM on Management of Data, 2023, 1 (2): 199.
6	CAO W, LIU Z J, WANG P, et al.. PolarFS: An ultra-low latency and failure resilient distributed file system for shared storage cloud database. Proceedings of the VLDB Endowment, 2018, 11 (12): 1849- 1862.
7	YANG Z K, YANG C H, HAN F S, et al.. OceanBase: A 707 million tpmC distributed relational database system. Proceedings of the VLDB Endowment, 2022, 15 (12): 3385- 3397.
8	VERBITSKI A, GUPTA A, SAHA D, et al. Amazon aurora: Design considerations for high throughput cloud-native relational databases [C]// Proceedings of the 2017 ACM International Conference on Management of Data. 2017: 1041-1052.
9	VERBITSKI A, GUPTA A, SAHA D, et al. Amazon aurora: On avoiding distributed consensus for I/Os, commits, and membership changes [C]// Proceedings of the 2018 International Conference on Management of Data. 2018: 789-796.
10	YANG J C, RAE I, XU J, et al.. F1 Lightning: HTAP as a service. Proceedings of the VLDB Endowment, 2020, 13 (12): 3313- 3325.
11	张超, 李国良, 冯建华, 等.. HTAP数据库关键技术综述. 软件学报, 2023, 34 (2): 761- 785.
12	ÖZCAN F, TIAN Y Y, TÖZÜN P. Hybrid transactional/analytical processing: A survey [C]// Proceedings of the 2017 ACM International Conference on Management of Data. 2017: 1771-1775.
13	KRASKA T, HENTSCHEL M, ALONSO G, et al.. Consistency rationing in the cloud: Pay only when it matters. Proceedings of the VLDB Endowment, 2009, 2 (1): 253- 264.
14	CHEN J, DING Y, LIU Y, et al.. ByteHTAP: Bytedance’s HTAP system with high data freshness and strong data consistency. Proceedings of the VLDB Endowment, 2022, 15 (12): 3411- 3424.
15	LU Y, LU Y, JIANG H. Adaptive consistency guarantees for large-scale replicated services [C]// Proceedings of the 2008 International Conference on Networking, Architecture, and Storage. 2008: 89-96.
16	YU H, VAHDAT A.. Design and evaluation of a conit-based continuous consistency model for replicated services. ACM Transactions on Computer Systems, 2002, 20 (3): 239- 282.
17	GAO L, DAHLIN M, NAYATE A, et al. Application specific data replication for edge services [C]// Proceedings of the 12th International Conference on World Wide Web. 2003: 449-460.
18	杨志丰. 分布式存储系统的一致性是什么? [EB/OL]. (2022-06-27)[2023-05-11]. https://zhuanlan.zhihu.com/p/34656939.
19	YAN H, SHAWN T. 使用TiDB读取TiFlash [EB/OL]. (2022-12-06)[2023-06-04]. https://docs.pingcap.com/zh/tidb/stable/use-tidb-to-read-tiflash.
20	CAO W, LI F, HUANG G, et al. PolarDB-X: An elastic distributed relational database for cloud-native applications [C]// Proceedings of the 2022 IEEE 38th International Conference on Data Engineering. 2022: 2859-2872.
21	SHEN S, CHEN R, CHEN H, et al. Retrofitting high availability mechanism to tame hybrid transaction/analytical processing [C]// Proceedings of the 15th USENIX Symposium on Operating Systems Design and Implementation. 2021: 219-238.

相关工作类别		连续一致性的支持	一致性变化方式	一致性种类	相关工作
分布式数据库复制服务		支持	中间件	连续一致性	文献[15-16]
云服务		不支持	自适应	线性一致、可串行	文献[13,17]
HTAP 数据库	OceanBase	不支持	系统变量、HINT	强一致、弱一致读	文献[18]
	TiDB	不支持	智能选择、HINT	线性一致读 (选节点读)	文献[19]
	PolarDB	不支持	HINT	最终一致、会话一致	文献[20]

查询	同步切换延迟/ms	查询	同步切换延迟/ms	查询	同步切换延迟/ms
Q1	12.50	Q9	12.20	Q16	3.85
Q2	4.54	Q10	2.39	Q17	5.51
Q3	15.80	Q11	3.00	Q18	4.19
Q4	2.86	Q12	15.50	Q19	5.97
Q5	3.69	Q13	4.36	Q20	2.54
Q6	7.75	Q14	3.67	Q21	13.00
Q7	4.84	Q15	6.12	Q22	16.30
Q8	4.44

查询	同步切换延迟/ms	查询	同步切换延迟/ms	查询	同步切换延迟/ms
Q1	0.001180	Q9	0.001750	Q16	0.001330
Q2	0.001040	Q10	0.002150	Q17	0.001340
Q3	0.001430	Q11	0.001540	Q18	0.001470
Q4	0.001140	Q12	0.001050	Q19	0.002940
Q5	0.001320	Q13	0.001140	Q20	0.002390
Q6	0.002560	Q14	0.000935	Q21	0.002620
Q7	0.000873	Q15	0.001240	Q22	0.001260
Q8	0.002050

实验组	查询负载描述	期望的切换效果	实际的切换效果
1	Q1—Q22全部在执行至第20秒时进行切换	切换	切换
2	除Q1, Q3, Q9, Q12, Q21, Q22几个同步代价大于10 ms的查询, 其余查询在执行至第20秒时切换	切换	切换
3	仅Q1, Q2这两个平均同步代价之和小于切换对齐代价的查询进行切换	不切换	不切换

实验组	查询负载描述	期望的切换效果	实际的切换效果
1	Q1—Q22全部以顺序一致性的新鲜度要求访问顺序一致性同步的分片	不切换	不切换
2	Q1—Q22查询负载的一致性需求在执行至第20秒时切换	切换	切换
3	除Q11—Q22查询外, 其余查询在执行至第20秒时切换	切换	切换
4	仅Q1, Q2这两个查询负载的一致性需求在执行至第20秒时发生改变	不切换	不切换