Distributed ID

Catalog

One 、UUID

Two 、 Database autoincrement sequence

3、 ... and 、redis、zk

Four 、 Snowflake algorithm

---

One 、UUID

1. Current date and time --- Time stamp
2. The time series --- Counter
3. Globally unique IEEE Machine identification number , If there is a network card , From the network card MAC The address for

advantage ：

1. The code is simple , Good performance （ Locally generated , No network consumption ）, Ensure that the only （ Relatively speaking ）

shortcoming ：

1. Every time I generate ID It's all out of order , And not all numbers , And there is no guarantee that the trend will increase ;UUID The result is a string , Poor string storage performance , Query efficiency is slow ;UUID It's too long , Not suitable for storage , Consume database performance ;ID No business implications , Poor readability ; There are information security issues , It may leak MAC Address .

Two 、 Database autoincrement sequence

standalone mode ：

advantage ：

1. Implement a simple , Rely on the database , The cost is small .
2. ID Digitization , Monotonic self increasing , Meet database storage and query performance .
3. It has certain business readability .（ Combined business code）

shortcoming ：

1. Strong dependence DB, There is a single problem , If the database goes down , The service is not available .
2. DB Generate ID Limited performance , Single point database is under great pressure , Unable to carry high concurrency scenarios .
3. Information security issues , For example, expose the order quantity ,url Change the query id Find other people's orders .

Database high availability ： Multi master mode for load , Start value and step size setting based on sequence , Different initial values , Same step size , The step size is greater than the number of nodes .

advantage ：

1. It's solved ID Generated single point problem , At the same time, the load is balanced .

shortcoming ：

1. System expansion is difficult , After the system defines the step size , Difficulty in adjusting step size after adding machine .
2. Database pressure , Get one at a time ID Must read and write the database once .
3. When the master and slave are synchronized , This data cannot be found due to data synchronization delay .

3、 ... and 、redis、zk

be based on redis Atomic operation of

be based on zk Of ZNode node

Four 、 Snowflake algorithm

64bit Integer number of , The first sign bit is fixed to 0,41 A time stamp ,10 position workId,12 A serial number . The number of bits can be implemented differently .

advantage ：

1. Each millisecond value contains ID It's worth a lot of , Not enough, you can change the number of digits to increase , Good performance （ rely on workId Realization ）.
2. The time stamp is high , In the middle is a fixed machine code , The self increasing sequence is in the low order , Whole ID It's an increasing trend .
3. Flexible layout of database nodes according to business scenarios bit Bit Division , High flexibility .

shortcoming ：

1. Strongly dependent on machine clock , If the clock goes back , Will lead to repeated ID Generate , Therefore, the algorithm based on this is generally used to find clock callback , Will throw exception handling , prevent ID Generate , This may cause the service to become unavailable .