An Tmplementation of Impure Suprogates

Surrogates or internal identifiers can be made to facilitate both fast access and storage independence if they are implemented properly. Such an implementation is discussed here; it permits the tuplcs of a relation to be accessed very fast by primary key for both random and sequential search without retarding the performance of secondary keys. It employs a key compression and a hasiiing algorithm and attempts to place tuples on data pages in the primary key sequence. Subsequent updates are absorbed by a dynamic allocation of overflows. An indexing technique called a hash tree holds surrogates in primary key order, and facilitates fast sequential access. The access speed remains high even at a 90% load factor, without being significantly affected by storage reorganisation resulting from the addition of new attributes, deletion of old attributes or change of data page sizes. These techniques have been implemented in the PRECI database system at Aberdeen.