Code that relies on the order of columns is asking for trouble anyway. It should never be a problem, since as a DBA, you're profiling and identifying SELECT *
statement, and ostracizing the perpetrators, correct?
When I do have to perform such an operation, due to some dumb legacy software, I would follow these steps:
- rename existing table and named constraints
- create table with new schema
- bulk insert from old to new table. if required, break on the id and do it in batches of 100,000* records using a while loop (* - a number that suits the hardware & environment)
- drop old table - reclaim space here
- create indexes with proper fillfactors
- create constraints (after indexes)
As you perform inserts updates and deletes, your indexes will become fragmented both internally and externally.
Internal fragmentation is you have a high percentage of free space on your index pages, meaning that SQL Server needs to read more pages when scanning the index.
External fragmentation is when the pages of the index are not in order any more, so SQL Server has to do more work, especially in IO terms to read the index.
If your indexes become too fragmented, at best, your queries will be less efficient but at worst, SQL Server will just stop using the indexes all together, meaning virtually all queries would have to perform a table scan or clustered index scan. This will hurt your performance a lot!
When you reorganise an index, then SQL Server uses the existing index pages and just shuffles data around on those ages. This will alleviate internal fragmentation and can also remove a small amount of external fragmentation. It is a lighter weight operation than rebuild and is always online.
When you rebuild an index, SQL Server actually resorts the data of the index and uses a new set of index pages. This will obviously alleviate both internal and external fragmentation but is a more heavy weight operation and by default causes the index to go offline, although it can be performed as an online operation, depending on your SQL Server version and settings.
Please do not expect to have 0 fragmentation after a Rebuild however. Unless you use a MAXDOP query hint, SQL Server will parallelise the rebuild operation and the more processors involved, the more fragmentation there is likely to be, because each processor or core, will rebuild their section or fragment of the index individually, without regard for each other. This is a trade off between best fragmentation levels and time taken to rebuild the index. For near 0 fragmentation, use MAXDOP 1 and sort the results in TempDB.
Best Answer
If you are just migrating the Schema itself, and not any of the data, Index Fragmentation of your user database tables should be irrelevant. The instructions in the migration script will come from system database tables holding the metadata for your user database/schema.