IO_STALL 问题和理解

Question

我每 5 分钟从 sys.dm_io_virtual_file_stats 收集 IO_STALLS，然后进行增量以查看哪些文件受 IO 影响最大。

在一个 5 分钟的时间内，我得到了 5826331 毫秒的增量，即 97 分钟。

我对此有点困惑，这是说 97 分钟前开始的操作仅在那时完成并因此记录了等待时间？

谢谢

根据要求添加代码：

/*

USE [SysDBA]
GO
*/
/****** Object:  Table [dbo].[DISKIOPS]    Script Date: 04/07/2013 11:40:15 ******/
/*
DROP TABLE [dbo].[DISKIOPS]
GO
*/
--Create the table
/****** Object:  Table [dbo].[DISKIOPS]    Script Date: 04/07/2013 11:40:15 ******/
/*
SET ANSI_NULLS ON
GO

SET QUOTED_IDENTIFIER ON
GO

SET ANSI_PADDING ON
GO

CREATE TABLE [dbo].[DISKIOPS](
    [IO_STALL] [bigint] NULL,
    [IO_STALL_READ_MS] [bigint] NULL,
    [IO_STALL_WRITE_MS] [bigint] NULL,
    [NUM_OF_READS] [bigint] NULL,
    [NUM_OF_WRITES] [bigint] NULL,
    [SIZE_ON_DISK_MB] [bigint] NULL,
    [DBNAME] [varchar](max) NULL,
    [NAME] [varchar](max) NULL,
    [FILE_ID] [int] NULL,
    [DB_FILE_TYPE] [varchar](max) NULL,
    [DISK] [varchar](max) NULL,
    [FILE_LOCATION] [varchar](max) NULL,
    [TIMESTAMP] [datetime] NULL
) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]

GO

SET ANSI_PADDING OFF
GO

*/

--Capture IO information from DMV and query to find deltas over time.
/*
USE [SysDBA]
GO

INSERT INTO [dbo].[DISKIOPS]
           ([IO_STALL]
           ,[IO_STALL_READ_MS]
           ,[IO_STALL_WRITE_MS]
           ,[NUM_OF_READS]
           ,[NUM_OF_WRITES]
           ,[SIZE_ON_DISK_MB]
           ,[DBNAME]
           ,[NAME]
           ,[FILE_ID]
           ,[DB_FILE_TYPE]
           ,[DISK]
           ,[FILE_LOCATION]
           ,[TIMESTAMP])
SELECT a.io_stall, a.io_stall_read_ms, a.io_stall_write_ms, a.num_of_reads, 
a.num_of_writes, 
--a.sample_ms, a.num_of_bytes_read, a.num_of_bytes_written,
( ( a.size_on_disk_bytes / 1024 ) / 1024.0 ) AS size_on_disk_mb, 
db_name(a.database_id) AS dbname, 
b.name, a.file_id, 
db_file_type = CASE 
                   WHEN a.file_id = 2 THEN 'Log' 
                   ELSE 'Data' 
                   END, 
UPPER(SUBSTRING(b.physical_name, 1, 2)) AS disk_location,
b.physical_name AS File_location,
GETDATE() AS Timestamp
FROM sys.dm_io_virtual_file_stats (NULL, NULL) a 
JOIN sys.master_files b ON a.file_id = b.file_id 
AND a.database_id = b.database_id
GO
*/
DECLARE @File_Name VARCHAR(8000),
        @Disk VARCHAR(5)
SET @File_Name = 'DBTEST'
SET @Disk = 'I:'
--Code to pull out deltas between collected IO stats.
;WITH IOPS   ([IO_STALL]
           ,[IO_STALL_READ_MS]
           ,[IO_STALL_WRITE_MS]
           ,[NUM_OF_READS]
           ,[NUM_OF_WRITES]
           ,[SIZE_ON_DISK_MB]
           ,[DBNAME]
           ,[NAME]
           ,[FILE_ID]
           ,[DB_FILE_TYPE]
           ,[DISK]
           ,[FILE_LOCATION]
           ,[TIMESTAMP]
           ,[ROW])
AS
(
SELECT *, ROW_NUMBER() OVER (PARTITION BY FILE_LOCATION ORDER BY TIMESTAMP DESC) AS [ROW]
FROM dbo.DISKIOPS 
)

SELECT MAX([IO2].[IO_STALL] - [IO1].[IO_STALL])
FROM IOPS IO1 JOIN IOPS IO2 ON IO1.ROW = (IO2.ROW+1)
WHERE IO1.NAME = IO2.NAME
AND IO1.Disk = @Disk

Answer 1

问题评论粘贴在下面：

io_stall本身并不意味着太多。如果在 10 秒内您有 1000 次操作每个暂停 1 秒，那么您将有 1000 秒的暂停。这将是 10 秒内超过 16 分钟的停顿。您需要将其与 IO 操作相关联...

上面是一个很好的例子，说明了如何看到不朽的和看似夸张的数字。就其本身而言，io_stall并不意味着什么。您需要知道该累积停顿的 I/O 操作规模。

而不是有这个：

SELECT MAX([IO2].[IO_STALL] - [IO1].[IO_STALL])
FROM IOPS IO1 JOIN IOPS IO2 ON IO1.ROW = (IO2.ROW+1)
...

您需要将停顿除以 I/O 操作以获得每个 I/O（或每个读取、或写入，或您正在寻找的任何粒度）的平均停顿。换句话说，我的建议是将您的查询修改为如下所示：

SELECT
    MAX(([IO2].[IO_STALL] - [IO1].[IO_STALL]) / (IO2.NUM_OF_READS + IO2.NUM_OF_WRITES - IO1.NUM_OF_READS - IO1.NUM_OF_WRITES))
FROM IOPS IO1 JOIN IOPS IO2 ON IO1.ROW = (IO2.ROW+1)

然后你需要有一个额外的谓词子句来确保你没有被零除：

...
WHERE IO1.NAME = IO2.NAME
and (IO2.NUM_OF_READS + IO2.NUM_OF_WRITES - IO1.NUM_OF_READS - IO1.NUM_OF_WRITES) > 0
AND IO1.Disk = @Disk

这基本上是计算每个 I/O 操作的平均值io_stall。就其本身而言，高io_stall可能只是意味着更高的工作量，而不一定是问题的迹象。