764 Project ideas

Generating a query plan that considers the existing bitmap and projection indexes. Consider different kinds of query plans such as left-deep or right-deep query trees. This involves coming up with cost formulas for the different alternatives and incorporating these into the optimizer so that a good plan is chosen.

This set of projects involves implementing various operators for executing star joins. This could be using bitmap join indexes with fact file scan or right deep query trees. There are some other combinations possible such as combining bitmap indexes with projection indexes. More thought is needed on this.

Given a table create a bitmap index for that. Explore efficient ways of doing this. Some possibilities are: building few bitmaps (as many can be held in memory) at a time and using multiple scans, successive refinement of equiwidth, creating bitmaps in segments using single scan (might lead to unclustered indexes).

Bitmaps can be stored in compressed or uncompressed format depending on selectivity. The operations should take care of these different formats for the input bitmaps. There should be a way for dynamically selecting the representation for the output bitmap depending on the operation and the expected selectivity of the output.

This involves rewriting the queries to make use of materialized views. The optimizer should consider alternative forms for the query and chooses a form with the least expected cost.

This project involves developing a frontend for the DBA in order to manage the creation of indexes, estimation of sizes, selection of views, precomputing them and identifying query profiles.

Study the type of queries an OLAP user might require. Consider some multidimensional data model and see if it is useful. Write to tool to finally convert the requests into SQL queries.

This project involves studying the OLAP council specification for multidimensional API and implementing it on top of the Paradise cursor.

Fact File is an efficient file organization supporting skipped sequential access. It uses a extent based tree with leaf extents chained. This project requires to come up with a CC / Recovery protocol and implement it.

Currently, fact file provides a single dimensional array like interface. Using this interface, we can read, update and delete nth record. Upon deletion, we allow the existence of holes. This project requires to implement an efficient mechanism for managing and reusing the holes. A related component associated with this project is appending records to fact file.

Studying JDBC specification and alternative methods of implementing them either map the calls to Java native methods on top of Paradise cursor class or make changes on the server side so that it directly returns data in the format that Java uses. Implement the best approach.

Histograms are used to estimate the result size of a query. They aid the optimizer to chose the cheapest plan. This project aims to incorporate histograms into Paradise optimzer. You have to implement both the statistics collection and incorporate it into the optimizer.