Massive data sets consisting of several million sequences provide the basis for new insights in a range of disciplines of medical and biological research. In order to realise their full potential, the evaluation process has to be backed up with an efficient bioinformatical processing chain including tasks such as clustering similar sequences. However, current exact solutions often constitute a bottleneck since they are usually base on pairwise sequence comparisons and, thus, are too slow to process data sets of today's magnitude within a reasonable time frame. Other approaches employ heuristics which are sufficiently fast but can introduce undesirable methodological errors in the data analysis. Hence, new strategies are needed in order to circumvent above problems. In this talk we address some related ideas which have been implemented in my group and which can used by Scientists for their research.