SequoiaDB V3.4 version of the official release, the database giant sequoias

Late autumn, SequoiaDB database giant sequoias in late autumn to give us "a fire." SequoiaDB v3.4 released it!

Distributed Transaction scene performance boost

SequoiaDB Sequoia database version 3.4 officially released, v3.4 of the most important features is to enhance the performance of distributed transaction scene. A comparison of the large version, SequoiaDB v3.4 scene in a distributed transaction, read and write performance improvements of up to 30% performance increase updated 1 times -1.5 times the query performance than the v3.2 upgrade more than 1.5 times.

Performance comparison a schematic old and new versions

For distributed transaction scenario, 3.4 version upgrade their technology mainly in the following: 

  • Improved 2PC Algorithm 

Distributed Intelligent arbitration algorithm affairs. 2PC distributed transaction is submitted to arbitration algorithm to increase intelligence, focused on solving the 2PC algorithm "In-doubt Transaction" abnormal state, to achieve the database in an extreme scenario for the multi-partition transaction intelligent arbitration to ensure distributed transaction strong consistency. 

  • Latch-less Memory Model 

To achieve multi-level memory pool and no lock memory model. Database cluster resource pool of memory, multi-level memory pool management, session access memory access achieved 99.99% hit rate, lock-free access memory under high concurrency OLTP scenarios, the system CPU usage increased by 10%. SequoiaDB v3.4 also provides online and offline monitoring memory memory analysis capabilities, automated generation memory analysis report. 

  • Improved Raft Algorithm

Raft breakthrough algorithm limits, to achieve full concurrency synchronization. SequoiaDB v3.4 introduced conflict arbitration mechanisms breakthrough synchronous serial plight only when there is a unique key constraint Raft algorithm to achieve synchronization between concurrent copy full record level, significantly increasing the efficiency synchronous copy. 

  • Improved Full-text Search Algorithm 

Full-text indexing performance optimization. Optimizing full-text indexing connection model, reducing connection time and memory usage, count operation of the full-text index hit directly by index results, significantly increasing the count data read performance.

目前,巨杉数据库针对金融交易场景,巨杉数据库已经规模应用在金融客户的核心交易、核心下移、关系型数据库替换等场景中,应用业务包括信用卡、网银、贷款等,在金融交易场景的应用,领先业界新一代分布式数据库。

SequoiaDB v3.4 功能提升

这次新版本全面提升金融级交易场景功能与性能,在分布式事务、数据一致性,并发CURD性能以及SQL兼容能力方面都做了深度优化。另外,为了满足金融级交易场景对稳定性严苛的技术要求,SequoiaDB 还升级了混沌测试框架,集群稳定性得到极大提升。
SequoiaDB v3.4的其他主要更新项如下:

存储引擎

  • 事务Auto-commit下推优化,将事务二阶段提交简化为一阶段提交,提升事务性能
  • 事务一致性确认机制
  • 实现多层级内存池和无锁内存模型
  • 全并发同步,提升副本数据同步性能
  • 提供增量数据归档、同步能力
  • 通过开启日志的全量模式和时间模式,可以实现按天,或指定时间对增量数据进行抽取,转换和归档,并将增量数据导入到其它ODS系统。
  • 全文索引支持数组类型
  • 全文索引支持 $or 和 $not 操作
  • 全文索引性能大幅提升
  • 访问计划增加自动过期清理,并实现对 $in 操作的参数化缓存能力
  • 插入数据支持重复键替代
  • 索引支持 not null 约束
  • 优化事务监控性能,实现无锁事务监控机制,减少事务监控管理对外部业务的性能影响

SQL引擎

  • 优化高可用能力,实现SQL引擎横向扩容
  • 算子下推存储节点,精确计算,提升网络带宽利用率
  • 事务Auto-commit下推存储引擎,简化事务二阶段提交为一阶段提交,提升事务性能
  • 支持NO TRANSACTION模式,提升初始化数据场景性能
  • 优化DDL操作,包括rename table,modify field,add primary key、index等操作
  • 全面兼容 MariaDB 语法

大对象引擎

  • 提供S3兼容的对象存储接口
  • 大对象存储支持按时间序进行自动分区,提升对大对象的存取和管理能力,可以快速按时间进行归档和清理
  • 大对象过滤支持过滤条件和精准匹配

易用性

  • 支持指定节点的重新选举能力
  • 提供 SQL 语法查询数据库当前状态与监控信息
  • 提供性能监控和慢查询分析能力

易用性进一步提高,巨杉工具矩阵正式推出

伴随 SequoiaDB v3.4 的发布,巨杉数据库也于近期推出了新一代的巨杉数据库工具矩阵,并计划在近期发布 SequoiaPerf 性能诊断工具。

SequoiaDB 工具矩阵示意图

SequoiaPerf 性能工具 即将发布

丰富的运维管理工具,是用户实使用数据库最直观的友好感受,巨杉数据库也将持续保持创新,聆听社区用户的声音,将面向用户的开发、运维、性能调优工具和最佳实践做到最好。

目前,巨杉数据库大型银行客户已经突破 50 家,应用场景包括核心交易、数据中台、内容管理和实时数据服务等。未来,我们也将保持自研和创新,在分布式数据库技术和多种行业应用中,保持领先。

Guess you like

Origin www.oschina.net/news/111499/sequoiadb-3-4-released