Column store database system in the market already invented by Vertica and then followed by SQLServer and Oracle. Last contender is MemSQL and MariaDB which is using MySQL client as the connector. They promise data compression and analytical query performance. Here we are try to load all dataset from this post into MemSQL finding the speed of both query.
Loading all files to in MemSQL take very significant time, 22 hours. This is possibly caused by times needed to compress the dataset. Loading process is using JDBC PreparedStatement with batch load for each file from host Max OSX and is takes only 3.85 GB disk storage. MemSQL is installed in VirtualBox Linux Centos 7 as required by MemSQL.
Compare to parquet here is the interesting result:
|Q1||1.136 sec||0.02 sec|
|Q2||1.442 sec||3.76 sec|
Here we can take conclusion that the query speed is depend on query load.