Table of contents
1. Preface
2. Selection dimension
3. Related information:
1. Preface
ArchitectAt work, you often encounter the problem of database storage selection, and there are many database products on the market, so you often have no idea how to start. Sometimes, the underlying database will be replaced many times during business development to online operation and maintenance, bringing unnecessary extra work to the entire R&D center.
When selecting database storage in combination with business scenarios, you can consider the following dimensions:
2. Selection dimension
1. Does the business need to ensure ACID transactions?
Transactions have four characteristics, namely atomicity, consistency, isolation and persistence.
YES: Mainstream relational database:MySQL, pg (PostgreSQL) etc.
NO: Mainstream NOSQL such as: MongoDB, hbase, cassandra, etc.
2. Require low latency?
The ultimate performance requirement and no need to worry about data loss: memcache
The data life cycle is short, the data volume is relatively small, the performance requirements are high, and the storage needs flexibleData structure：Redis
General scyladb>canssdra>hbase
3. Is there a high concurrent write, and there are more random reads than range reads?
YES: hbase, cassandra, hbase has better reading range.
4. Massive data storage, support horizontal expansion, distributed fault tolerance?
YES: Distributed NOSQL such as: resis, hbase, Cassandra, mongodb
5. Need a full text search?
YES: Used in massive dataElasticsearch; The data volume can be controlled using pg GIST index.
6. Store complex data models?
YES: Document database mongodb
7. OLTP OR OLAP OR offline analysis?
OLTP (online transaction processing): row storage database mysql, pg
OLAP (online analysis processing): column database high compression gp, clickhouse
Offline analysis: hbase
8. Financial level requires high availability and data loss
oceanbase, Alibaba Cloud mysql financial version, mysql group replication
9. Need transactions and massive data storage
Library and table:drds、sharding-jdbc+Relational database
newsql：TiDB
10. CAP tradeoffs
CAP： Consistency, Availability, Partition tolerance.
CA: Relational database
CP ：hbase、redis、mongodb
AP ：cassandra、riak 、couchdb
11. Special business scenarios
Geographic Information System GIS: PG>ES>Mongodb
Timing: TaosDB>influxdb>opentsdb
12. Development perspective, business iteration is fast, and schema free is needed
Development-friendly: Relational databases are OK. Another is MongoDB, but it also has two sides. It is not as strict as relational databases. Inserting any value of any type into any set can sometimes bring negative effects.
13. Operation and maintenance angle
Easy Operation and Maintenance cassandra>hbase
3. Related information:
Alibaba Cloud Developer Community—Database Selection:What should be considered for database selection? -Q&A-Ali Cloud Developer Community-Ali Cloud
What is ClickHouse:/likun557/article/details/109733541
The advantages and disadvantages and application scenarios of four popular databases, MongoDB, ElasticSearch, Redis, and HBase:The advantages and disadvantages and application scenarios of four popular databases, MongoDB, ElasticSearch, Redis, and HBase - Zhihu