Integer identities can be useful to a database internally, but perhaps you shouldn't be using an identifying key externally, especially if you have multiple web servers or expose the IDs to users. An alternative is to use a Globally Unique Identifier (GUID), referred to as UNIQUEIDENTIFIER in SQL Server. We have already touched on these, as they are generated by the NEWID() function used in the suboptimal sampling example.
GUIDs are used ubiquitously and are 16 bytes long, four times bigger than an integer. The size of GUIDs means that you are unlikely to get a unique constraint conflict when inserting a random GUID into an already populated table.
People sometimes worry about GUID collisions, but the numbers involved are so staggeringly ...
