1.1 A Brief History of Distributed Database

1.1 A Brief History of Distributed Database （分布式数据库发展简史）

主讲人：

Liyuan Feng(冯立元)

SQL Engine of PingCAP

Qiniu Cloud / PingCAP

Distributed storage / Database engineer

Old days (20th century)

• Traditional Database
  • Interact with SQL
  • Single node / Shared everything
  • OLTP workloads

PostgreSQL、MySQL、SQL Server

单机数据库，OLTP，数据量少，业务并发不高。

Internet decades(2000s)

• Distributed middleware

ShardingSphere、Vitess

业务增多，分库分表。

• The rise of Data Warehouses
  • Interact with programming language API

Hadoop

数据量暴增，大数据兴起，不支持 SQL 交互，采用基于编程语言的计算框架。

• The rise of Data Warehouses
  • Interact with SQL
  • Distributed / Shared nothing
  • Column-oriented
  • OLAP workloads

Greenplum、monetdb

OLAP 兴起，大多是分布式的，按链存储。

• NoSQL
  • Interact with key-value API
  • No transaction, No relation model, No Schema
  • High performance
  • High Availability
  • Distributed / Shared nothing

MongoDB、HBase、Cassandra

放弃关系型模型和事务，高性能，高可用，分布式。

NewSQL(2010s)

• NewSQL provides the scalability of NoSQL systems for OLTP workloads while not giving up ACID transactions
  • Interact with SQL
  • Distributed / Shared nothing

Cloude Spanner

HTAP(2010s)

• HTAP
  • Execute OLTP queries like NewSQL
  • Execute OLAP queries like a data warehouse

混合事务分析处理（OLTP + OLAP）

And TiDB tries to bring the best to you

• An HTAP NewSQL database
  • Interact with SQL
  • OLTP & OLAP workloads
  • Distributed / Shared nothing
  • High performance
  • High Availability