As most developers are used to working with relational databases, the typical path to failure starts with a data model that closely resembles one found in a RDBMS. A loading job is then written, which only succeeds in crashing their nodes every couple of hours.
Finally, once the data is there, they build their application around a RDBMS framework driver (because their team has experience with a specific framework) that was augmented to (rather than designed to) work with Cassandra.
In the end, the app team blames the DBAs for an application with high latency, a trainwreck of a data loader, and request timeouts about as common as firecrackers in the U.S. in early July.
All of these issues are preventable. What's more, responsibility ...
