HP Big Data Analytics: Effects of Compression on Query Performance

Name: Kirby Sand
E-mail: kirby.sand@hp.com
Knowledge Required: Ideal candidates have a strong knowledge of Linux operation systems. Candidates should also have knowledge of databases and how they work. Knowledge of networking is a bonus.
Motivation: Our Big Data system must process and store 350GB/day of data. It is critical that we effectively use compression to reduce the storage footprint while maintaining query performance.
Description: PageWide Web Press (PWP www.hp.com/go/pagewidewebpress), a printing division within HP Corvallis, has a large scale database extensively used for troubleshooting product issues and doing business analytics. Our products produce 350GB of data per day and create database tables with billions of rows. We are looking to examine the effectiveness of different compression algorithms on storage and query performance using an Oracle database. This project will take a fundamental look at storage size and query performance while implementing a variety of compression techniques. You will measure CPU usage, Disk I/O, Access Path, Memory, Time, etc. to determine the most performant data storage techniques.
Objectives: The objective of the project will be to find a solution that minimizes our storage footprint while not compromising (or potentially improving) query performance. Oracle provides a lot of different ways to compress your data. We will explore them all.
Deliverables: OSU students will provide the following:
• Background research, form hypotheses, document expected outcome
• Design and run experiments
• Collect the data from the experiments
• Propose and implement available compression techniques in Oracle
• Benchmark results
• Written analysis of the findings
Other comments: This project will be research based. It will be the foundation of a white paper, and will be presented at Hotsos (the leading Oracle Performance Tuning conference in the United States).

   D. Kevin McGrath
   Last modified: Mon Oct 16 16:07:06 2017