The Columbia Fast Query Project  

---

The aim of this project is to make queries on databases fast, in the face of:

• Huge data sets, typical of data warehouse applications
• Complex queries
• Sophisticated machines (with complex and sometimes delicate performance parameters) running the database code.

• Databases on Multi-Threaded, Multi-Core machines.

Power consumption issues are forcing CPU chip designers to increase performance in new ways. Rather than simply increasing clock speeds, they are placing multiple CPU cores on a single chip, and are providing hardware to run multiple logical threads on each CPU. Running a database system on such an architecture can lead to interference, because the CPUs on a chip and the threads within each CPU share (and thus compete for) resources. We are examining ways to reduce or eliminate such interference for query intensive workloads.
 
• Architecture-Sensitive Databases.

Now that very large databases can fit in the RAM of cheap machines, one can develop database techniques that perform well in that context.  For example, one might try to design query processing algorithms that yield good data reference locality, and hence low CPU cache miss rates.  Other architectural features are branch-misprediction effects and the availablilty of SIMD instructions (such as SSE on x86 platforms).
 
• Materialized Views.

One can potentially answer queries faster by using stored answers to commonly-used query subexpressions.  These stored answers are called materialized views.
 
• Query Processing Algorithms and Query Optimization Techniques.

By developing new algorithms for common and/or critical database operations, one can improve query performance. Choosing a good evaluation plan for a complex query is a difficult problem.  New optimization strategies could help find better plans.

Available Software.

• Database Parallelism on Multicore Machines
• Le Subscribe

More information can be found in our publications.

Database Research Group

This material is based in part upon work supported by the National Science Foundation under Grants IRI-9457613, IIS-9812014, IIS-0120939, and IIS-0534389.
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.