I currently use uuid family key project, each time is disordered, the system impact on the performance of the future? B is, after all, a tree structure, index up bad lane, and used to support the prefix index? Going to do an identification number, from upstream to downstream, identification number will be more and more big, such as I somewhere from head office warehouse stock warehouse, the warehouse of the object code is: supplier identification code/batch code/head office warehouse code/branch warehouse/classification yard/objects on the code, how can I design makes this code is a globally unique but not too long?