SQL情况 <Value> 在SQL Server 2005而不是2000上运行缓慢

Question

I have a very simple query that is driving me absolutely nuts.

Here's the situation:

• I have two database servers.
• One is an old SQL Server 2000 (VM), very minimal resources.
• The other is a very large SQL Server 2005 enterprise cluster with an absolutely ridiculous amount of resources available.
• I have a small portion of a larger query that executes in 3 seconds and returns 50,000+ rows of data on SQL Server 2000
• This same small query takes upwards of 15 minutes to return 1000 rows on SQL Server 2005
• The database I'm working with is a mirror image on these two servers. Same tables, same data in tables, same indexes on tables, etc.

I've tried creating different indexes on the SQL Server 2005 tables, defragged all the indexes that exist, updated table statistics, etc. Nothing has been able to make this query run faster on SQL Server 2005. Currently nothing else is running against the SQL Server 2005 server, and our DBAs assure me it's not a configuration issue or anything to do with feature deprecation between SQL Server 2000 and SQL Server 2005.

The query is below:

SELECT (CASE 
             WHEN TeamMember.ID IN  (SELECT DISTINCT ProjMgrID FROM ProjMgr)
                THEN 'Yes' 
                ELSE 'No' 
        END) AS OnProjAsMgr 
FROM TeamMember

So return a distinct list of all the ProjMgrs, and if the TeamMember is on that list then assign "Yes" to the OnProjAsMgr value.

I'm a total SQL newbie, and this is code written by a predecessor. I don't know if there's a better way to write it, but I cannot figure out why it runs great on SQL Server 2000 but completely implodes on SQL Server 2005.

Answer 1

DISTINCT causes a sort, and IN causes the entire query to be evaluated. How does this version work:

SELECT OnProjAsMgr = CASE WHEN EXISTS 
  (SELECT 1 FROM dbo.ProjMgr WHERE ProjMgrID = TeamMember.ID)
  THEN 'Yes' 
  ELSE 'No' 
  END
FROM dbo.TeamMember;

If that fares no better then I suspect indexes are missing and no query will perform well without them.

Answer 2

I don't know about "better" way, but here's a different way that you can try:

SELECT ... other fields you want ...,
       CASE WHEN ProjMgr.ProjMgrID IS NULL
            THEN 'No'
            ELSE 'Yes'
        END AS OnProjAsMgr
  FROM TeamMember
  LEFT
 OUTER
  JOIN ProjMgr
    ON ProjMgr.ProjMgrID = TeamMember.ID
;

Notes:

• On some DBMSes, this will perform much better than your query, but I don't know whether it would perform better on SQL Server 2005, given that your query already works fine on SQL Server 2000. I think you'll just have to try it.
• Your query uses SELECT DISTINCT ProjMgrID FROM ProjMgr . If that is actually different from SELECT ProjMgrID FROM ProjMgr — that is, if there are actually duplicate values of ProjMgrID in ProjMgr — then the above query is not exactly equivalent to yours, since it will give a separate record for each duplicate. You might have to add a GROUP BY clause in that case.
• For that matter, if SELECT DISTINCT ProjMgrID FROM ProjMgr is equivalent to SELECT ProjMgrID FROM ProjMgr for your data, as I suspect, then it also might be worth removing the DISTINCT , since that might affect the optimizer as well.

Answer 3

SELECT (
    CASE WHEN P.ID IS NULL THEN 'No' ELSE 'Yes' END
) AS OnProjAsMgr
FROM TeamMember AS T 
LEFT OUTER JOIN ProjMgr AS P ON T.ID=P.ID