The DCB-Tree: A Space-Efficient Delta Coded Cache Conscious B-Tree

Main-memory index structures have become mainstream for a large number of problem domains. However, in the case of web-based datasets, which feature exponential growth, it is an ongoing challenge to fit those data entirely in main-memory. In this paper, we present the DCB-Tree, an extremely space efficient main-memory index structure for the storage of short fixed-size keys. It features a two-stage cache-line aligned node layout. In comparison to other main-memory index structures it reduces the amount of memory required by 80% in the best and by 30% in the worst case. Although it is tailored towards space consumption, it features good overall performance characteristics. In particular, in the case of very large real world datasets it provides performance equal or superior to state of the art main-memory index structures.