Modeling splits in file structures (Q1105353)

We analyze the expected behaviour of file structures where splits are used to handle overflows. Two cases are analyzed. The first model is of a file with an index on top of the data structure. We analyze the effect of unbalanced splits, and the effect of splitting in more than two buckets. The second model is of an ideal hash file, in wich the probability of insertion remains the same for every bucket, regardless of how many times the bucket has been split. The result is an upper bound in any dynamic hashing method that uses splitting and does not allow overflow records. In both cases, the effect of using partial expansions is included.