Why data isolation

Someone put Saas Compared with the traditional software service mode, it is ： Well drilling and tap water , Drilling a well is a one-time payment. You can use it whatever you want in the future , For tap water, you can pay as much as you use , Not so free .

Again ： It turns out that you can't eat food in winter , Each family has its own pickle urn , This urn is only used in winter , Waste space . Someone started building a place to store pickles , The pickle container must distinguish which is from which .Saas When the product is sharing micro Services , Each tenant's own data cannot be shared , Need to be isolated .

Isolation plan

The storage data is mainly relational database , other NoSql The database is a functional auxiliary , Can be divided into ：

1. sub-treasury ： Different tenants use different libraries

2. table ： Different tenants use different tables

3. Add... To the table TenantID Distinguish

4. A mixed solution

Isolate database ： sub-treasury

Applicable scenario ： Independent deployment requirements , Big customers

advantage ： The coding is simple , Single tenant project system development is the same ; Database performance optimization can be differentiated according to different schemes of tenants ; Data backup is easy ;

shortcoming ： The cost is relatively high , The database needs to be deployed separately .

Technical realization ：Sping Medium AbstractRoutingDataSource;Mycat middleware . Specific implementation reference ：

middleware .Mycat First experience _ Idle cat's blog -CSDN Blog

Spring. Dynamic data sources AbstractRoutingDataSource_ Idle cat's blog -CSDN Blog

Shared database , Isolated data structure ： table

Different tenants use different tables , Of course, there are business dimensions . such as ： Student list , The teacher table

Tenant 1：student_1, teacher_1

Tenant 2：student_2, teacher_2

advantage ： The cost of warehouse distribution is too high , Split tables are convenient for both optimization and backup , And it doesn't cost that much

shortcoming ： To create a tenant, you need to generate a table, that is DDL operation , Such as ： according to student Of Schema establish student_1 surface , It can be more complicated in function ：1. When is the table created ;2. Which tables to create ;3. How to distinguish different tables when adding functions ; 4. Base surface Schema How to store All need to be designed . The reference scheme is ： When designing functional permissions , Associate tables with functions , Create the tables involved when creating a new tenant , Find the mapping table according to the opened function .

difficulty ： Developers don't know when writing code ID,sql I can only write student, And actually implemented sql Need modification ：student_1, Reference plan ：Saas. Data access control （Sql analysis ）_ Idle cat's blog -CSDN Blog

Expand ：student and teacher It's all watches , Whether it can also be abstracted as a table ,root The table shows that there are only id,updateby, updatetime ,createby ,createtime ,isdeleted ,remark And other system fields ,student Table is business key by “student” Of root Sub table of ,root_student_1 （ Tenant 1 The student watch of ）

Applicable scenario ： Generally not used alone , Does not conform to the progressive law of product development ( The development cost is relatively high , At the beginning, the project resources are insufficient ), The actual development is to develop the single system first , The multi tenant transformation will be carried out after the business is verified , Therefore, the cost of sub warehouse at this stage is the lowest .

Shared data structure ： Add... To the table TenantID

Scheme description

It's easy to say , Any system seems to add one tenant_id( Tenant id) It becomes saas The system . For example, the original user login is ：

select * from student where email='[email protected]'

Change to

select * from student where email='[email protected]' and tenant_id =1;

For complex business saas System , This is very dangerous , And the development efficiency is very low . Think about it. If that programmer writes sql When did you forget to add “ and tenant_id =1” . The result is unimaginable .

scene ： It is suitable for scenarios with small amount of data , Small and medium-sized enterprises (smes) , Get customer scenarios for free

advantage ： No additional resources are required , You only need to add one field , And then modify Sql Just go

shortcoming ： There is no way to optimize according to the tenant , Backing up data by tenant is complex , Fault isolation is difficult

difficulty ： The framework level needs to be addressed tenant_id Condition add as well as insert When value Add data in . Refer to the post ：Saas. Data access control （Sql analysis ）_ Idle cat's blog -CSDN Blog

A mixed solution

A system is relatively large , Basically, not everything will follow one plan . such as ：

1. Login logs with small amount of data There is no need to divide the library , table . Add one tenantid It's easy to solve

2. A huge number of logs , Operation behavior table , Must be divided into tables , You may have to divide the table from multiple dimensions , Such as ： Period of time , Tenant ID

3. Different business scenarios ： Generally, it will be split according to groups , For their own simple maintenance , It's all sub warehouse . Each library will be divided into tables or added according to the actual situation tenantid programme

4. Expand ： Metadata driven scheme , The above schemes are from ： Business , Operation and maintenance , cost , Development and other considerations , Easy to think of , If it is Saas It can be supported in the middle and early stages , But if you want to answer KA Big customers , For example, there will be characteristic development , Manifested as ： Extend the original functions, such as table fields , There are also completely customized pages . How make ？ Metadata , stay 【 table . Expand 】 One column and then continue to abstract , To put it simply, put Mysql Medium Schema Tables and data sheets , All moved to Saas Platform self-service control , This scheme is not just a data persistence scheme , It also needs to cover other needs , This is it. Saas The difficulties of , We should have a global vision , Also need to focus a little .

appendix

The term

DB：DataBase abbreviation , database

DS：DataSource abbreviation , data source

TenantID： Tenants in multi tenancy ID , Unique to identify a tenant

Schema： Definition , This article refers to the definition of data table

Lazy loading ： Load only when using resources