Paper Abstract

We tested Cloudera Impala in an effort to understand what hardware setup would provide the best performance/price for it. Our aim is to provide a quick practical guide for choosing the infrastructure to run Impala on.

What's Inside

1. Query Execution Times
   Results for a set of 20 queries run on the same data set, ten times on each hardware configuration.
2. Single vs. Dual-CPU Instances
   We weren't expecting dramatic score changes between single and dual-CPU instances, but what we found was surprising.
3. Performance/Price Scores
   Using our standard cost structure, we added the price per hour for each instance and paired it with the Impala performance score.