For a primary key of an integral type, why is it important to avoid gaps?

Question

I am generating a surrogate key for a table & due to my hi/lo algorithm, everytime you reboot/restart the machine, gaps may appear.

T1: current hi = 10000000         


        

Answer 1

why is it important that there be no gaps in the sequences

It is not important. Gaps are fine. For performance reasons, gaps are tolerated.

What would be useful is a guarantee to have an strictly increasing sequence (i.e. the sequence has the same ordering as the row creation time). But even that is not guaranteed in a clustered configuration (with local counter caches).

Answer 2

From a primary key or indexing internals perspective, why is it important that there be no gaps in the sequences?

It's not important - what lead you to believe it was?
All that matters with the primary key is that it is unique to all the data in the table. Doesn't matter what the value is, or if the records before and after are sequencial values.

Answer 3

Who told you this? A surrogate key has no meaning at all, so there can't be any gap. What is a gap in something that has no meaning? We use UUID's for our keys, something like this: 6ba7b812-9dad-11d1-80b4-00c04fd430c8. What would be the "next" key? Nobody knows, nobody cares. As long as it is unique, it's fine.