A high-level comparison of several SQL-on-Hadoop engines plus a classic SQL database server.
Operational processing is a use case where zero-latency data is presented to users. It usually involves a data source that is continuously updated with new data and concurrently queried. Seconds and even microseconds count in operational processing. It could be the difference between success and failure. Operational reporting is the basis for popular trends such as operational intelligence or real-time analytics.
Interactive processing (OLAP, self-service BI, data visualization) is the classic use case of data warehouse environments. With interactive reporting, users repeatedly access their data to see what’s happening with certain business processes. They see their data organized as cubes, dimensions, and hierarchies. They can filter their data, drill down to a more detailed level or, vice versa, do a roll-up, and they can apply aggregations and statistical functions, such as sum, average, and standard deviation. It’s called interactive because the users are continuously sending new queries to the database.
Batch processing is the most classic use case of reporting in the IT industry. Usually, on predefined days and times reports are created and the results are distributed to the business users. The queries executed in batch reporting environments usually consist of simple joins and no complex analytical functions. This use case fits like a glove for SQL-on-Hadoop. Almost all these engines can support this type of workload.