Int Versus Char Primay Key Performance
Jul 23, 2005
Hi,
My company has a scenario where we would like to change the data type
of an existing primary key from an integer to a char, but we are
concerned about the performance implications of doing so. The script
for the two tables that we need to modify is listed below. Table
FR_Sessions contains a column named TransmissionID which is currently
an integer. This table contains about 1 million rows of data. Table
FR_VTracking table also contains the TransmissionID as part of it's
primary key and it contains about 35 millions rows of data. These two
tables are frequently joined on TransmissionID (FR_Sessions is the
parent). The TransmissionID column is used primarily for joins and is
not typically displayed.
We need like to change the TransmissionID data type from int to
char(7), and I had a few questions:
1) Would this introduce significant performance degradation? I have
read that char keys/indexes are slower than int/numeric.
2) Are there collation options (or any other optimizations) that we
could use to minimize the performance hit of the char(7)...if so which
ones?
I am a software architect by trade, not a database guru, so please go
easy on my if I overlooked something obvious :)
Any suggestions or information would be greatly appreciated.
Thanks,
Tim
-------------------
CREATE TABLE [FR_Sessions] (
[TransmissionID] [int] IDENTITY (1, 1) NOT NULL ,
[PTUID] [varchar] (10) COLLATE SQL_Latin1_General_CP1_CI_AS NOT NULL ,
[PortNum] [numeric](6, 0) NOT NULL CONSTRAINT [DF_FR_Sessions_PortNum]
DEFAULT (0),
[CloseStatus] [varchar] (20) COLLATE SQL_Latin1_General_CP1_CI_AS NULL
,
[RecvBytes] [int] NULL ,
[SendBytes] [int] NULL ,
[EndDT] [datetime] NULL CONSTRAINT [DF_FR_Sessions_EndDT] DEFAULT
(getutcdate()),
[LocalEndDT] [datetime] NULL ,
[TotalTime] [int] NULL ,
[OffenderID] [numeric](9, 0) NULL ,
[UploadStatus] [char] (1) COLLATE SQL_Latin1_General_CP1_CI_AS NOT
NULL CONSTRAINT [DF_FR_Sessions_UploadStatus] DEFAULT ('N'),
[SchedBatchID] [numeric](18, 0) NULL ,
[SWVersion] [varchar] (10) COLLATE SQL_Latin1_General_CP1_CI_AS NULL ,
[DLST] [bit] NULL ,
[TZO] [smallint] NULL ,
[Processed] [bit] NOT NULL CONSTRAINT [DF_FR_Sessions_Processed]
DEFAULT (0),
[CallerID] [varchar] (13) COLLATE SQL_Latin1_General_CP1_CI_AS NULL ,
[PeerIP] [varchar] (16) COLLATE SQL_Latin1_General_CP1_CI_AS NULL ,
[XtraInfo] [varchar] (1024) COLLATE SQL_Latin1_General_CP1_CI_AS NULL
,
[IdType] [char] (1) COLLATE SQL_Latin1_General_CP1_CI_AS NULL ,
CONSTRAINT [PK_FR_Sessions] PRIMARY KEY CLUSTERED
(
[TransmissionID]
) WITH FILLFACTOR = 90 ON [PRIMARY]
) ON [PRIMARY]
CREATE TABLE [FR_VTracking] (
[TransmissionID] [int] NOT NULL ,
[FrameNum] [int] NOT NULL ,
[LatDegrees] [float] NOT NULL ,
[LonDegrees] [float] NOT NULL ,
[Altitude] [float] NOT NULL ,
[Velocity] [float] NOT NULL ,
[NumPositions] [smallint] NOT NULL ,
[NavMode] [smallint] NOT NULL ,
[Units] [smallint] NOT NULL ,
[GPSTrackingID] [numeric](18, 0) IDENTITY (1, 1) NOT NULL ,
[dtStamp] [datetime] NULL ,
CONSTRAINT [PK_FR_VTracking] PRIMARY KEY CLUSTERED
(
[TransmissionID],
[FrameNum]
) WITH FILLFACTOR = 90 ON [PRIMARY]
) ON [PRIMARY]
View 5 Replies
May 15, 2008
I was hoping I wouldn't be another poster with performance issues after migrating to SQl 2005 from SQL 2000 but here I am.
I am in the process of testing out our databases on Sql Server 2005 for migration from SQL Server 2000 and there are certain portions of code that have been affected negatively. I have read thru many of the posts here and have tried out most of the recommendations. I will start out with things I've done and then provide the actual SQL.
1) I have rebuilt all indexes ( using the DBCC REINDEX using the table option).
2) Updated the db engine to latest hot fix (build 3239) that addresses speed related fixes.
3) I also ran sp_createstats using the 'fullscan' option to create stats on all columns of all tables (minus indexed columns)
4) Since nothing seemed to work, I even ran UPDATE STATICS with FULL SCAN on all tables even though I did not need it as the REBUILD woudl have created stats. But I was willing to try anything.
I have confirmed that the execution plans are different even though the data on both sql 2000 and sql 2005 are identical (i put a copy on 2005). The plans themselves are huge as the queries are huge. Here is the query.
SELECT InterimView.* ,TestView.*
FROM View_LabDataExport_TestFormData_55 TestView
RIGHT OUTER JOIN ( SELECT ReqView.*, CDView.*
FROM View_LabDataExport_FormData_55 ReqView
LEFT OUTER JOIN View_LabDataExport_FormData_CD_55 CDView
ON ( CDView.DB_SubjectID_CD = ReqView.DB_SUbjectID )
) InterimView
ON ( InterimView.DB_FormID = TestView.DB_FormID_T AND
InterimView.DB_LabSampleID = TestView.DB_LabSampleID_T )
The above query takes abotu 8 secs to run on 2000 and about 1 minute to run on 2005. This is for a small dataset and on larger datasets this is only going to more pronounced ( as confirmed by other teams that have already migrated in my company). Another point worth mentioning might be if I remove the TestView.* from the select list, it works in 5 to 6 seconds. Is there an issue with Sql 2005 and a large number of columns or anything of that sort? On 2000, the time remains the same , about 8 seconds if I remove this from the select list.
Here is the statistics ion on 2005
(21234 row(s) affected)
Table 'Worktable'. Scan count 75490, logical reads 3676867, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabTestToReportPanel'. Scan count 476, logical reads 1524, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabReportPanel'. Scan count 0, logical reads 260, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'DiscreteValue'. Scan count 1, logical reads 176106, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabReleasedSampleTest'. Scan count 1, logical reads 2078, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabSample'. Scan count 1360, logical reads 18567, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Form'. Scan count 2302, logical reads 8225, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabTest'. Scan count 1, logical reads 23, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabSampleDef'. Scan count 1, logical reads 10530, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabArea'. Scan count 1, logical reads 2, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Lab'. Scan count 1, logical reads 2, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Location'. Scan count 1, logical reads 2, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Study'. Scan count 0, logical reads 6, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Item'. Scan count 1335, logical reads 32940, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'ObjectState'. Scan count 1, logical reads 10972, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Object'. Scan count 0, logical reads 20674, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Subject'. Scan count 0, logical reads 3293, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'FormDef'. Scan count 2, logical reads 70, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'PrintedLabSampleLabel'. Scan count 0, logical reads 13144, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'PrintedForm'. Scan count 0, logical reads 4219, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'StudySite'. Scan count 0, logical reads 2756, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'StudyEvent'. Scan count 18, logical reads 40, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'StudyEventDef'. Scan count 0, logical reads 36, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'FormDefToStudyEventDef'. Scan count 1, logical reads 43, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'LabSampleDefToFormDef'. Scan count 1, logical reads 255, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Here is the statistics ion on 2000
Table 'LabTestToReportPanel'. Scan count 2123, logical reads 4820, physical reads 44, read-ahead reads 0.
Table 'LabReportPanel'. Scan count 130, logical reads 260, physical reads 0, read-ahead reads 0.
Table 'DiscreteValue'. Scan count 103914, logical reads 208214, physical reads 0, read-ahead reads 0.
Table 'Location'. Scan count 19031, logical reads 38062, physical reads 2, read-ahead reads 0.
Table 'Lab'. Scan count 19031, logical reads 38062, physical reads 0, read-ahead reads 0.
Table 'LabArea'. Scan count 19031, logical reads 38062, physical reads 0, read-ahead reads 0.
Table 'LabSampleDef'. Scan count 24670, logical reads 49340, physical reads 0, read-ahead reads 0.
Table 'LabTest'. Scan count 19406, logical reads 39575, physical reads 0, read-ahead reads 0.
Table 'LabReleasedSampleTest'. Scan count 4289, logical reads 73865, physical reads 1014, read-ahead reads 24.
Table 'Study'. Scan count 4291, logical reads 8582, physical reads 0, read-ahead reads 0.
Table 'LabSample'. Scan count 5647, logical reads 31382, physical reads 308, read-ahead reads 4.
Table 'Form'. Scan count 4291, logical reads 9272, physical reads 2, read-ahead reads 10.
Table 'PrintedLabSampleLabel'. Scan count 4289, logical reads 17097, physical reads 114, read-ahead reads 308.
Table 'ObjectState'. Scan count 6860, logical reads 13760, physical reads 1, read-ahead reads 0.
Table 'Object'. Scan count 6860, logical reads 23559, physical reads 90, read-ahead reads 701.
Table 'PrintedForm'. Scan count 1375, logical reads 4505, physical reads 40, read-ahead reads 16.
Table 'StudySite'. Scan count 1378, logical reads 2756, physical reads 4, read-ahead reads 0.
Table 'Subject'. Scan count 1599, logical reads 3332, physical reads 2, read-ahead reads 0.
Table 'StudyEvent'. Scan count 18, logical reads 52, physical reads 0, read-ahead reads 0.
Table 'StudyEventDef'. Scan count 18, logical reads 54, physical reads 0, read-ahead reads 2.
Table 'FormDefToStudyEventDef'. Scan count 1, logical reads 69, physical reads 0, read-ahead reads 23.
Table 'FormDef'. Scan count 2, logical reads 78, physical reads 1, read-ahead reads 4.
Table 'LabSampleDefToFormDef'. Scan count 1, logical reads 308, physical reads 1, read-ahead reads 306.
Table 'Item'. Scan count 1335, logical reads 36510, physical reads 140, read-ahead reads 1047.
(21234 row(s) affected)
(147 row(s) affected)
One difference between the two is the work table that 2005 creates versus 2000. I can attach the plans but they are huge. I will attach it if you ask.
What I was looking for was suggestions on what I could do short of rewriting code or any suggestions in general.
FYI, this has also been posted on the SQL Server Engine forum.
Thanks
View 10 Replies
View Related