Clickhouse learning (I) Clickhouse?

---

What is? ClickHouse

Is a columnar storage oriented database , And allow to use SQL Query and generate analysis report in real time , It is mainly used for OLAP( Online analysis processing query ), It has not gone in the field of big data Hadoop ecology , instead Local attached storage As the storage , So the whole IO Maybe not Hadoop The limitations of that set .

characteristic

• It is the fastest for single table query

In the field of in memory database, it is known as the fastest

• Diversity engine

Contains the merge tree 、 journal 、 Interfaces and other four categories 20 Multiple engines

• Have DBMS( Database management system ) The function of

Almost covered the standard SQL Most of the grammar of , Include DDL and DML, And various supporting functions , User management
Management and authority management , Data backup and recovery

• High throughput write capability

It's just working LSM Tree The structure and Hbase equally , Because this structure is written in sequence when importing data , The written data cannot be changed ,so The background will merge regularly , According to official disclosure benchmark The test shows that 50MB-200MB/s Write throughput of , According to each line 100Byte Estimate , About equal to 50W-200W strip /s Write speed of

• Data partitioning and thread level parallelism

Partition management of data avoids the problem of full table query , And through multiple CPU The core processes part of it separately to realize parallel data processing , This leads to , Single Query You can use the whole machine CPU

shortcoming

• Not friendly enough for concurrency support . In high concurrency queries , If the number of queries at the same time is about thousands , It's easy to have problems

• Relative to query , His writing speed is not so excellent , The write speed of a single machine can only reach about 30W/s～60W/s

• Compared with his own single table query performance ,join The performance of is not very excellent