Optimizing Memory-Resident Decision Support System Workloads for Cache Memories

In the second part of this work cache optimizations are proposed for two database system components: algorithms and query optimizer. In the former, blocking and prefetching are applied to database algorithms. In the latter the first public domain cache-oriented query optimizer is proposed. This optimizer chooses the ordering of operations and implementation of those operations using the number of cache misses and the number of instructions as the metric. In an evaluation of the proposed optimizations using a real architecture, some complex queries show performance improvement over the existing optimizations. One query from a standard benchmark achieves 29% improvement while the average for five queries is 13%. While this improvement is moderate, these optimizations are implemented in the database system without any changes to the hardware. Therefore the proposed optimizations provide improvement at no additional cost. A sensitivity test showed that the improvement provided by the proposed optimizations is independent of changes to the cache configuration like cache size, line size, and miss penalty. Finally, the prefetching optimization doubles the performance improvement from 13% to 28% in average for all queries.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Trancoso, Pedro P.M.

Contributors dc:contributor

• Torrellas, Josep

• Computer Science

Language dc:language

eng

Identifier

(MiAaPQ)AAI9912399

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/81936