Monday, February 9, 2015

Hadoop based SQL engines

Apache Hadoop comes more and more into the focus of business critical architectures and applications. Naturally SQL based solutions are the first to get considered, but the market is evolving and new tools are coming up, but leaving unnoticed.

Listed below an overview over currently available Hadoop based SQL technologies. The must haves are:
Open Source (various contributors), low-latency querying possible, supporting CRUD (mostly!) and statements like CREATE, INSERT INTO, SELECT * FROM (limit..), UPDATE Table SET A1=2 WHERE, DELETE, and DROP TABLE.

Apache Hive (SQL-like, with interactive SQL (Stinger)
Apache Drill (ANSI SQL support)
Apache Spark (Spark SQL, queries only, add data via Hive, RDD or Parquet)
Apache Phoenix (built atop Apache HBase, lacks full transaction support, relational operators and some built-in functions)
Presto from Facebook (can query Hive, Cassandra, relational DBs & etc. Doesn't seem to be designed for low-latency responses across small clusters, or support UPDATE operations. It is optimized for data warehousing or analytics¹)
VoltDB (ACID compatible, ANSI SQL near 92, low-latency multi query engine)
SQL-Hadoop via MapR community edition (seems to be a packaging of Hive, HP Vertica, SparkSQL, Drill and a native ODBC wrapper)
Apache Kylin from Ebay (provides an SQL interface and multi-dimensional analysis [OLAP], "… offers ANSI SQL on Hadoop and supports most ANSI SQL query functions". It depends on HDFS, MapReduce, Hive and HBase; and seems targeted at very large data-sets though maintains low query latency)
Apache Tajo (ANSI/ISO SQL standard compliance with JDBC driver support [benchmarks against Hive and Impala])
Cascading's Lingual² ("Lingual provides JDBC Drivers, a SQL command shell, and a catalog manager for publishing files [or any resource] as schemas and tables.")

Non Open Source, but also interesting
Splice Machine (Standard ANSI SQL, Transactional Integrity)
Pivotal Hawq (via Pivotal HD, ANSI SQL 92, 99 and OLAP)
Cloudera Impala (SQL-like, ANSI 92 compliant, MPP and low-latency)
Impala does not incorporate usage of Hadoop, but leverages the cached data of HDFS on each node to quickly return data (w/o performing Map/Reduce jobs). Thus, overhead related to performing a Map/Reduce job is short-cut and one can gain improvements runtime.
Conclusion: Impala does not replace Hive. However, it is good for different kind of jobs, such as small ad-hoc queries well-suited for analyzing data as business analysts. Robust jobs such as typical ETL taks, on the other hand, require Hive due to the fact that a failure of one job can be very costly.

Thanks to Samuel Marks, who posted originally on the hive user mailing list