April 4, 2014 | Robert Wilkinson
Open-Source SQL-on-Hadoop Query Engines in Benchmark
Today we are excited to release a new benchmark that looks at the performance comparison of leading open-source SQL-on-Hadoop query engines.
The benchmark and analysis used distributions from Hortonworks, Cloudera, and InfiniDB that included Hive 0.11/0.12, Impala 1.2, Presto 0.57, and InifiniDB for Hadoop 4.0.
In the benchmark, InfiniDB performed extremely well. This is one of those places where smaller numbers are better - i.e. less time (smaller number) = faster (better performance).
For example, in the snapshot below from Query 8 of 10, you can barely see InfiniDB. It is up to 80x faster than competing solutions.
Highlights of the benchmark and testing include:
- The InfiniDB engine delivered up to 40x performance gains over Hive on reporting queries and ran ad-hoc and analytic queries “interactively” that were not run to completion on Hive due to the excessive time required.
- The InfiniDB engine delivered 20-40x performance advantage over Presto. Presto was the runner-up finisher in terms of query coverage and able to complete 9 of 10 queries.
- The InfiniDB engine performed similarly to Impala for reporting queries, but showed significant performance gains on ad-hoc queries and was able to run analytic queries that are unsupported by Impala.
- InfiniDB was the only SQL engine that completed all reporting, adhoc, and analytic queries in the benchmark, and was the fastest in 6 out of 10 queries.
You can get your free complimentary copy of the report here (login required).
We are pretty excited about how InfiniDB continues to scream past the competition in testing and is proven to be a quick time-to-value solution because it is SQL accessible for directly querying HDFS.
Let us know how InfiniDB is helping you conquer your Big Data problems.