Automated Index Tuning


A database management system provides an excellent means to organize a large amount of data and access it quickly. In some cases, it is only possible to maintain acceptable performance if an expert system administrator spends time tuning the system, i.e., selecting features and configuring parameters in a way that improves the overall system performance. A crucial step in the tuning process is the selection of indexes that allow important portions of the data to be found quickly. Index selection is often a hard problem, requiring extensive knowledge of the database system's internals and the applications that access the database.

This research project aims to develop tools in order to make index tuning easier and more effective. Our work has focused on online techniques that monitor the actual queries submitted to the system and automatically choose indexes that are expected to improve the performance of the current workload. The techniques that we have developed have been integrated in the PostgreSQL database management system.

The project is funded by a grant from the Los Alamos National Laboratories, by NSF award IIS-1018914, by an AWS in Education research grant, and by gifts from IBM, NEC and Oracle.


External Collaborators

Past Collaborators


  1. Jeff LeFevre, Jagan Sankaranarayanan, Hakan Hacigümüs, Jun'ichi Tatemura,  Neoklis Polyzotis: "Exploiting Opportunistic Physical Design in Large-scale Data Analytics". CoRR (2013).
  2. Hakan Hacigümüs, Jagan Sankaranarayanan, Jun'ichi Tatemura, Jeff LeFevre, Neoklis Polyzotis: "Odyssey: A Multi-Store System for Evolutionary Analytics". PVLDB6(11): 1180-1181 (2013)
  3. J. LeFevre, J. Sankaranarayanan, H. Hacigumus, J. Tatemura, N. Polyzotis, "Towards a Workload for Evolutionary Analytics", In DanaC 2013 (Data Analytics in the Cloud)
  4. R.Wang, Q.T.Tran, I.Jimenez, N.Polyzotis, "INUM+: A leaner, more accurate and more efficient fast what-if optimizer", in SMDB 2013
  5. Karl Schnaitter, Neoklis Polyzotis, "Semi-Automatic Index Tuning: Keeping DBAs in the Loop", to appear in VLDB 2012 (Full version as CoRR abs/1004.1249).
  6. M. Consens, K. Ioannidou, J. LeFevre, N. Polyzotis, "Divergent Physical Design Tuning for Replicated Databases", to appear in SIGMOD, 2012.
  7. I. Jimenez,  H. Sanchez, Q.T. Trung, N. Polyzotis, "Kaizen: A Semi-Automatic Index Advisor", to appear in SIGMOD (Demonstration track), 2012 .
  8. Ivo Jimenez, Jeff LeFevre, Neoklis Polyzotis, Huascar Sanchez, Karl Schnaitter,  "Benchmarking Online Index-Tuning Algorithms", IEEE Data Eng. Bull. 34(4): 28-35, 2011.
  9. D. Dash, N. Polyzotis, A. Ailamaki, "CoPhy: A Scalable, Portable, and Interactive Index Advisor for Large Workloads", in VLDB 2011.
  10. Ioannis Alagiannis, Debabrata Dash, Karl Schnaitter, Anastasia Ailamaki, Neoklis Polyzotis, "An automated, yet interactive and portable DB designer", In SIGMOD (Demonstration Track), pp. 1183-1186, 2010.
  11. K. Schnaitter, N. Polyzotis, and L. Getoor, "Index Interactions in Physical Design Tuning: Modeling, Analysis, and Applications", In Proceedings of the VLDB Endowment, 2(1), 2009, pp. 1234-1245.
  12. K. Schnaitter and N. Polyzotis, "A Benchmark for Online Index Selection", In ICDE 2009, pp. 1701-1708.
  13. Karl Schnaitter, Serge Abiteboul, Tova Milo, Neoklis Polyzotis, "On-Line Index Selection for Shifting Workloads", In ICDE Workshops, pp. 459-468, 2007 (Slides)
  14. Karl Schnaitter, Serge Abiteboul, Tova Milo, Neoklis Polyzotis, "COLT: continuous on-line tuning", in SIGMOD (Demonstration Track), pp. 793-795, 2006. (Slides)