Updates and deletes does not necessarily acquire an exclusive lock from the start. Sometimes it might be shared locks (or update locks) competing with exclusive locks causing the problems and...

- Keeping the non-unique clustered index would be painful as SQL would have to increase the index size with its unique identifier.- Keeping the non-unique clustered index would be painful...

Just remember that CXPacket waits are often a side effect of inefficient queries or indexes leading to unnecessary CPU usage. If you run large DW queries, it might not be...

Thanks! Perfect..

One thing to keep in mind is fragmentation. When new ID's are increasing at the end, page splits will be far less than trying to insert a value randomly in...

Hi Jason,

I am curious, why don't you suggest mirroring for this?

Thanks

Row Level Locking is the default on an instance, database and the objects in the DB.

It can be disabled on an object level, and you can see if it is...

Also: "share a number of fields" let the alarm bells go off. Also not sure what you want, but your database is not normalized. SQL is pretty easy if your...

Is it then perhaps timing out when the table is created? Do you see Sch-S (schema stability) locks appearing?

Grant Fritchey (4/24/2013)

---

It sounds like a blocking issue. You can use the software suggested or just look at sys.dm_exec_requests to see if there are processes holding locks that will prevent...

Bulk insert into a staging table and then building the clustered and other indexes afterwards will generally be faster.

This is also useful when you then want to switch the...

Agree. Simple method for a small resultset. Best performance is still a CLR based one.

One set based way of doing it:

DECLARE @Adding TABLE (ID int identity(1,1),Result int,ActualResult int)

INSERT INTO @Adding (Result) VALUES (10),(10),(10),(10),(-10),(-10),(-10),(-10)

SELECT ID, Result, (Select SUM(result) From @Adding A2 WHERE A2.Id <=...

I think you mean row compession.. Page compression will also include row compression.

I would also use a hashbyte function. In theory there still might be some duplicates, although unlikely.