Architecture: database architecture design

introduction

• This paper introduces the architecture design of database ;

• Usually , A single machine cannot meet the read and write requirements of large systems for databases , It must be solved by clustering ;

• The introduction of clusters means that the complexity of the system is increased , Make the system complex and difficult to maintain ;

• Database load balancing strategy is usually adopted 、 Read write separation strategy 、 Optimize the strategy of dividing database and table ;

Load balancing

• Extensibility is strong ： When the system needs higher database processing speed , As long as you simply add a database server, it can be expanded ;

• Maintainability ： When a node fails , The system will automatically detect the failure and transfer the application of the failed node , Ensure the continuous work of database ;

• Security ：

  • Because the data will be synchronized on multiple servers , Data set redundancy can be realized , Security through multiple data ;

  • Put the database into the intranet , Better protection of database security ;

• Ease of use ： Completely transparent to applications , What the cluster exposed is a IP(1) Can't follow Web The processing power of the server distributes the load ;

• shortcoming ： Load Balancer ( Control terminal ) fault , The whole database system will be paralyzed ;

Read / write separation

• When the frequency of reading and writing varies a lot （ebay The read-write ratio of is 260:1）, To improve efficiency , Reduce disk IO pressure , Adopt read-write separation ;

• Realization principle ：

  • Database server builds master-slave cluster , A master from 、 One master and many followers can ;

  • The database host is responsible for reading and writing operations , The slave machine is only responsible for reading operations ;

  • The database host synchronizes the data to the slave through replication , Each database server stores all the business data ;

  • The business server sends writes to the database host , Send the read operation to the database slave ;

• Solution of data inconsistency between master and slave ：

  • Data inconsistency ： When the data is written to the master server , Can't be found until the next synchronization ;

  • Read the host again after the slave fails ;

  • Key business ( account number 、 Transfer, etc ) All read and write operations point to the host , Non critical business adopts read-write separation ;

Sub database and sub table

Sub database

• It refers to splitting into different databases according to functional modules , For example, it can be divided into order warehouse 、 Commodity bank 、 User base ;

• join It is only applicable to the joint query of different tables in the same database , After splitting, different databases cannot be used join Statement to query , It can only be queried several times ;

• Transaction is a concept in the same database , To rollback transactions between different databases , Only query log The way to roll back ;

• The high cost , Splitting into different databases means that multiple backup databases need to be established ;

Sub database table - vertical ( The longitudinal ) Split

• Originally, a table contained all 10 A field ;

• Now separate the fields with high query frequency into another table （ For example, marriage website name, sex, age Three fields ）, Other fields （ Such as personal introduction destribution etc. ） Leave it on the original watch ;

• advantage ： Query performance improvement , If you only query important fields , There is no need to check out other fields , fast ;

• shortcoming ： If you want to find all the fields , It must be queried twice ;

Sub database table - level ( The transverse ) Split

• Save the data of the same table in blocks to different databases , The tables in these databases have exactly the same structure ;

• Sequential routing ：

  • For example, it can be divided by the year of the date of the order ,2003 Years of putting db1 in ,2004 Year of db2, And so on ;

  • The disadvantage is uneven data distribution , Probably 2003 Orders for 100W,2008 Year old 500W;

• hash route ：

  • Yes user_id Conduct hash（ Or if user_id If it's numerical, use it directly user_id The value is also acceptable. ）, And then use a specific number , For example, a database needs to be divided into 4 Database , We will use 4 That's right user_id Of hash Value for modulo , Decide which table to store ;

  • If you want to 4 A table becomes 5 Tables , Change the membrane value , Then all the data needs to be changed , so much trouble ;

• Configure the routing ：

  • It's about building a DB, This DB Save separately user_id To DB The mapping relation of , Every time you visit the database, you need to query the database first , To get specific DB Information , Then we can perform the query operations we need ;

  • advantage ： Strong flexibility , One to one relationship ;

  • shortcoming ： One more query before each query , Will cause certain performance loss ;

source: //changsiyuan.github.io/2018/07/17/2018-7-17-Architecture-6

Share & Looking at