No SQL in Practice: A Write-Heavy Enterprise Application

The continuous information growth in current organizations has created a need for adaptation and innovation in the field of data storage. Alternative technologies such as NoSQL have been heralded as the solution to the ever-growing data requirements of the corporate world, but these claims have not been backed by many real world studies. Current benchmarks evaluate database performance by executing specific queries over mostly synthetic data. These artificial scenarios, then, prevent us from easily drawing conclusions for the real world and appropriately characterize the performance of databases in a real system. To counter this, we used a real world enterprise system with real corporate data to evaluate the performance characteristics of popular NoSQL databases and compare them to SQL counterparts. In particular, we present one of the first write-heavy evaluations using enterprise software and big data. We tested Cassandra, Mongo DB, Couchbase Server and MS SQL Server and compared their performance while handling demanding and large write requests from a real company with an electrical measurement enterprise system.