A systematic Benchmarking on the performance of Spark-SQL for processing Vast RDF datasets
This project is maintained by DataSystemsGroupUT
The following figures show the comparative representation of partitioning techniques (i.e. Horizontally, Subject-based, Predicate-based) for 100M, 250M, and 500M respectively.
*Please note that the following ranking figures include Hive as a 5^th storage backend. However, we have excluded Hive and for simpolicity kept only the HDFS storage file formats (CSV, AVro, ORC, and Parquet). Figures of ranking the partitoning dimension (excluding Hive) can be shown here.