SQL Server 有序GUID,SequentialGuid,

问题描述


有序的GUID性能对比,堪比自增ID integer

 

一个大神告诉我NEWSEQUENTIALID() 在数据迁移的时候会有问题(感谢大神指点),所以我就深挖一下这个函数。

    关于NEWSEQUENTIALID() 的用法 参照  NEWSEQUENTIALID()

NEWSEQUENTIALID 是对 Windows UuidCreateSequential 函数的包装。

https://msdn.microsoft.com/zh-cn/library/ms189786(v=sql.120).aspx

我们系统中对UuidCreateSequential 方法的扩展是这样写的,代码如下:

 

    public static class GuidExtension
    {
        [DllImport("rpcrt4.dll", SetLastError = true)]
        public static extern int UuidCreateSequential(out Guid guid);
        private const int RPC_S_OK = 0;

        public static Guid CreateRpcrt4Guid()
        {
            Guid guid;
            int result = UuidCreateSequential(out guid);
            if (result == RPC_S_OK)
            {
                byte[] guidBytes = guid.ToByteArray();
                Array.Reverse(guidBytes, 0, 4);
                Array.Reverse(guidBytes, 4, 2);
                Array.Reverse(guidBytes, 6, 2);

                return new Guid(guidBytes);
            }
            else
                return Guid.NewGuid();
        }

    }

 

  有以下几个缺点:

  1、暴漏MAC地址:NEWSEQUENTIALID函数最后6个字符是网卡的MAC地址

  可以执行看一下

create table #t
(
    id uniqueidentifier not null  default newsequentialid()
    ,name varchar(100)
)
go

insert into #t(name)
output inserted.id
values('a')
  

 

  2、如果进行数据迁移,到另一台机器上,MAC地址改变就会引起页的争用。

    因为GUID在的SQL Server的值大小的比对是这样的:

with uids as (
            select id =  1, uuid = cast ('00000000-0000-0000-0000-010000000000' as uniqueidentifier)
    union   select id =  2, uuid = cast ('00000000-0000-0000-0000-000100000000' as uniqueidentifier)
    union   select id =  3, uuid = cast ('00000000-0000-0000-0000-000001000000' as uniqueidentifier)
    union   select id =  4, uuid = cast ('00000000-0000-0000-0000-000000010000' as uniqueidentifier)
    union   select id =  5, uuid = cast ('00000000-0000-0000-0000-000000000100' as uniqueidentifier)
    union   select id =  6, uuid = cast ('00000000-0000-0000-0000-000000000001' as uniqueidentifier)
    union   select id =  7, uuid = cast ('00000000-0000-0000-0100-000000000000' as uniqueidentifier)
    union   select id =  8, uuid = cast ('00000000-0000-0000-0010-000000000000' as uniqueidentifier)
    union   select id =  9, uuid = cast ('00000000-0000-0001-0000-000000000000' as uniqueidentifier)
    union   select id = 10, uuid = cast ('00000000-0000-0100-0000-000000000000' as uniqueidentifier)
    union   select id = 11, uuid = cast ('00000000-0001-0000-0000-000000000000' as uniqueidentifier)
    union   select id = 12, uuid = cast ('00000000-0100-0000-0000-000000000000' as uniqueidentifier)
    union   select id = 13, uuid = cast ('00000001-0000-0000-0000-000000000000' as uniqueidentifier)
    union   select id = 14, uuid = cast ('00000100-0000-0000-0000-000000000000' as uniqueidentifier)
    union   select id = 15, uuid = cast ('00010000-0000-0000-0000-000000000000' as uniqueidentifier)
    union   select id = 16, uuid = cast ('01000000-0000-0000-0000-000000000000' as uniqueidentifier)
)
select * from uids order by uuid desc

 

输出结果:

  类似 汉字的三点水偏旁(为了好记)


从这里可以看出,MAC地址对GUID的大小有这最高的决定性,这就导致在数据迁移的时候出问题。

 

COMB解决方案


 

 COMB 类型的GUID 基本设计思路是这样的:既然GUID数据生成是随机的造成索引效率低下,影响了系统的性能,那么能不能通过组合的方式,保留GUID的前10个字节,用后6个字节表示GUID生成的时间(DateTime),这样我们将时间信息与GUID组合起来,在保留GUID的唯一性的同时增加了有序性,以此来提高索引效率。

 

前十个字节是通过随机数生成

private static readonly RNGCryptoServiceProvider RandomGenerator = new RNGCryptoServiceProvider();

      byte[] randomBytes = new byte[10];
      RandomGenerator.GetBytes(randomBytes);

 

 

后六个字节用时间生成

      long timestamp = DateTime.UtcNow.Ticks / 10000L;
      byte[] timestampBytes = BitConverter.GetBytes(timestamp);

      if (BitConverter.IsLittleEndian)
      {
        Array.Reverse(timestampBytes);
      }

 

 

最后组合起来

    byte[] guidBytes = new byte[16];
    Buffer.BlockCopy(randomBytes, 0, guidBytes, 0, 10);
    Buffer.BlockCopy(timestampBytes, 2, guidBytes, 10, 6); 

    return new Guid(guidBytes);

 

 

这个解决方法是被大家所认可的,唯一感觉不好的地方是,在快速获取很多的GUID的时候,时间是一样的,加上随机生成的数据,这一组数据是大小不一的。假如数据库里有很多数据,这一组数据肯定比他们大,性能应该没有问题。

github地址:

https://github.com/nhibernate/nhibernate-core/blob/master/src/NHibernate/Id/GuidCombGenerator.cs#L25-L72

https://github.com/jhtodd/SequentialGuid/

 

我的解决方法


 

总结上面的方法,UuidCreateSequential 前面10个字节有序,后6个是MAC地址。COMBO解决方案是前面10个随机,后六个是时间。我是将这两个结合起来

前10个去UuidCreateSequential 方法的值,后6个取时间

代码:

public static Guid NewSequentialGuid()
    {
        const int RPC_S_OK = 0;
        Guid guid;
        int result = UuidCreateSequential(out  guid);

        if (result != RPC_S_OK)
        {
            throw new System.ComponentModel.Win32Exception(System.Runtime.InteropServices.Marshal.GetLastWin32Error());
        }
        else
        {
       //这里把UuidCreateSequential函数返回的数据做处理
byte[] guidBytes = guid.ToByteArray(); Array.Reverse(guidBytes, 0, 4); Array.Reverse(guidBytes, 4, 2); Array.Reverse(guidBytes, 6, 2);
       //这里用时间
long timestamp = DateTime.UtcNow.Ticks / 10000L; byte[] timestampBytes = BitConverter.GetBytes(timestamp); if (BitConverter.IsLittleEndian) { Array.Reverse(timestampBytes); }
       //最后把时间赋值给后6位 Buffer.BlockCopy(timestampBytes,
2, guidBytes, 10, 6); return new Guid(guidBytes); } } [System.Runtime.InteropServices.DllImport("rpcrt4.dll", SetLastError = true)] private static extern int UuidCreateSequential(out Guid guid);

 

这里可以在程序调用,作为DBA在数据库使用的话可以将这个方法添加到程序集里,需要有些改动

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Data.SqlTypes;

public class FunctionNewGuid
{
  //这里需要添加SqlFunction属性
  //返回类型是数据库类型
[Microsoft.SqlServer.Server.SqlFunction]
public static SqlGuid NewSequentialGuid() { const int RPC_S_OK = 0; Guid guid; int result = UuidCreateSequential(out guid); if (result != RPC_S_OK) { throw new System.ComponentModel.Win32Exception(System.Runtime.InteropServices.Marshal.GetLastWin32Error()); } else { byte[] guidBytes = guid.ToByteArray(); Array.Reverse(guidBytes, 0, 4); Array.Reverse(guidBytes, 4, 2); Array.Reverse(guidBytes, 6, 2); long timestamp = DateTime.UtcNow.Ticks / 10000L; byte[] timestampBytes = BitConverter.GetBytes(timestamp); if (BitConverter.IsLittleEndian) { Array.Reverse(timestampBytes); } Buffer.BlockCopy(timestampBytes, 2, guidBytes, 10, 6); return new SqlGuid(guidBytes); } } [System.Runtime.InteropServices.DllImport("rpcrt4.dll", SetLastError = true)] private static extern int UuidCreateSequential(out Guid guid); }

 

编译生成DLL后,注册到数据库

--设置数据库是可信任
 ALTER DATABASE TEST SET TRUSTWORTHY ON

 --创建程序集
 CREATE ASSEMBLY SQLCLR FROM 'D:\SQLCLR.DLL'
 WITH PERMISSION_SET = UNSAFE


 --用程序集方法创建函数
 CREATE FUNCTION func_NewSequentialGuid()
    RETURNS uniqueidentifier
AS external name SQLCLR.FunctionNewGuid.NewSequentialGuid

 

    

测试代码:

 批量请求:


 

select dbo.func_NewSequentialGuid() 
union
select dbo.func_NewSequentialGuid() 
union
select dbo.func_NewSequentialGuid() 
union
select dbo.func_NewSequentialGuid() 
union
select dbo.func_NewSequentialGuid() 

 结果:

 

 

 多次请求:


 

create table #t
(
    uuid uniqueidentifier 
    ,id int identity
)
go

insert into #t(uuid)
values(dbo.func_NewSequentialGuid())
go 10

select * from #t

 

 

 

git地址

https://gitee.com/wangzhanbo/cms/tree/master/Library

 

如果有问题,希望大家指正。。。

 

 

posted on 2018-04-17 14:04  王占波  阅读(3831)  评论(2编辑  收藏  举报

导航