Physical design in operational environments is based on workloads. Audit or log files are analyzed to find the most critical queries (either frequent, expensive, or with a bounded execution time). Based on these, we can decide what indexes or structures have to be defined.

Several wizards already exist in the market to help this process. Nevertheless, analytical environments are different in that a user does not usually repeat the same queries, all of them are relatively expensive, and their execution time is not bounded. If we want to avoid the intervention of the IT department in the tuning of the systems, an alternative would be to do it based on the collaborative experience of users.

Users should state their requirements and they should be cross-analyzed to detect similarities and incompatibilities. Then, structures can be proposed based on those conciliated requirements. This topic will study how this can be handled and to which extent it results in significant performance improvements. Special emphasis will be put in NOSQL storage, which are gaining more and more relevance as a mean to store Big Data, and how to improve their performance based on previous user stories.

Main Advisor at Universitat Politècnica de Catalunya (UPC)
Co-advisor at Aalborg Universitet (AAU)