应用错误收集

我正在进行一些性能测试，发现了一个我无法解释的结果。我以两种方式实现相同的逻辑：T-SQL UDF和.Net CLR功能。这是源代码：

标量UDF：

create or alter function dbo.ParticularDateFormat(@inputDate date)
returns char(14)
as
begin
   declare @return char(14)
   set @return = cast(DATEPART(dd, @inputdate) as char(2)) + ' ' + choose(datepart(mm, @inputDate), 'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dez') + ' ' + cast(DATEPART(yyyy, @inputDate) as char(4))
   return @return
end

.Net CLR功能

    [SqlFunction(IsDeterministic = true)]
    public static SqlString NewDateStyle(SqlDateTime inputDate)
    {
        return ((DateTime)inputDate).ToString("dd MMM yyyy");
    }

然后我在100万行表上运行这两个函数。当我运行标量UDF时，我得到这个统计时间反馈：

 SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 0 ms.
SQL Server parse and compile time: 
   CPU time = 0 ms, elapsed time = 0 ms.

(1000000 row(s) affected)
Table 'SmallTable'. Scan count 1, logical reads 1616, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

(1 row(s) affected)

 SQL Server Execution Times:
   CPU time = 3734 ms,  elapsed time = 4587 ms.

这是我运行.Net CLR功能时的统计时间反馈：


 SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 0 ms.
SQL Server parse and compile time: 
   CPU time = 141 ms, elapsed time = 191 ms.

(1000000 row(s) affected)
Table 'SmallTable'. Scan count 1, logical reads 1616, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

(1 row(s) affected)

 SQL Server Execution Times:
   CPU time = 1312 ms,  elapsed time = 3586 ms.
SQL Server parse and compile time: 
   CPU time = 0 ms, elapsed time = 0 ms.

 SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 0 ms.

正如预期的那样.Net CLR功能运行速度更快。

我无法理解的是经过时间过长的原因，几乎是处理查询时间的两倍（CPU时间）。

有什么想法吗？

.Net CLR函数和Scalar UDF性能问题

0 个答案: