Supervised by:

Indexing is a very important concept in database systems. Index structures are maintained to organize huge data collections in an efficient way. Many of these structures are tuned for the use on conventional hard drives. In this thesis we focus on indexing on solid state drives based on flash memory. Solid state drives have other characteristics than conventional hard drives. They do not contain any mechanical parts and the underlying NAND flash memory confers them other performance characteristics. The access latency is much lower and the write performance is slowed by the flash memory limitations. We built two different index structures: a conventional B+Tree and an index structure based on CSS-trees with an in-memory delta. On the B+Tree both drives had a constant performance. On the CSS-tree hierarchy the throughput varied depending on the number of trees in the list. In the experiments our solid state drive reached a much higher throughput (at least by a factor of 3) on both index structures. Comparing the two indexes the solid state drive favored the delta indexing approach. Additionally one observes the huge progress of this new technology. The density grows, the costs fall and the performanc