xsevensinzx - Wednesday, February 27, 2019 6:28 PM
Please don't misunderstand me. I have no problem with the use cases you mentioned nor do I have an issue with an SQL interface on these environments. The value in getting the data loaded and investigating what's there and what can be done is quite invaluable. The issue I have is when those techniques are implemented in a production environment in lieu of a relational database when the data being loaded is relational (most come from another RDBMS) and the tools are implemented to operate as though it were an RDBMS. It was set up like a Rube-Goldberg machine with a complex set of moving parts all to avoid using a database. That appeared to be a requirement as the conversations kept circling around their "Big Data" solution. The volume of data was not that large that it couldn't have easily set in a Postgres database (assuming they had to stay in the AWS space). The approach was decided upon long before any data analysis was performed. When the Enterprise Architect and I kept asking why we couldn't just use a database instead of the complicated process and we were told that our suggestion wasn't the "Big Data" way. Time to implementation wasn't a concern because it's taken over 2 years to implement the solution using their framework.
Clearly the platform was chosen long before any analysis was done to determine the correct solution. This leads to all sorts of problems. I have witnessed a project that failed miserably because SQL Server was used for document management instead of a more appropriate platform like MongoDB. Each platform has its strengths and weaknesses, what it does well and what it does not do well. Using the wrong tool for the job only invites mediocre results at best.
LinkedIn: http://www.linkedin.com/in/sqlrv
Twitter: http://www.twitter.com/sqlrv
Website: http://sqlrv.com