Research and Implementation of Index Scheme for Large-Scale Dataset

Based on the characteristics that B⁺-tree is not suitable for indexing large-scale dataset, which could cause random disk I/O increased and more pages split when a large number of random data is inserted, this paper presents a new data structure, which can not only index more than a billion records, but also has a higher efficiency when do random and sequential search. What´s more, the performance is relatively stable for large-scale data updating. Experiments show that the new data structure performs well and valid when we index large-scale dataset.