TPC-DS benchmark

2019-01-27 夜未深，，，今天给大家加个餐，，，貌似大家突然对TPC-DS的跑分有了兴趣。。。来分享一篇一年半前的旧文。。。https://databricks.com/blog/2017/07/12/benchmarking-big-data-sql-platforms-in-the-cloud.html Databricks对Spark做了调优，然后快了五倍。。。裸跑TPC-DS其实很没意思。。。

1）Spark on Databricks outperforms vanilla Spark on AWS by 5X using the same hardware specs.

2）Spark on Databricks outperforms Presto by 8X. While Presto could run only 62 out of 104 queries, Databricks ran all.

3）Spark on Databricks not only outperforms the on-premise Impala by 3X on the queries picked in the Cloudera report, but also benefits from S3 storage elasticity, compared to fixed-physical disks on-premise.