Although NoSQL databases tend to offer similar feature sets, the way they implement them varies substantially from one product to the next. Even in cases where systems choose the same consistency models, there are often very significant implementation differences and configurations that have a significant effect on how they can be configured and perform. This presentation will discuss:

• Tunable consistency and durability models in major NoSQL databases such as Cassandra, MongoDB, Couchbase, Aerospike, and others.
• Differences between how these databases achieve said models, and what it means in terms of application setup and hardware investments.
• The tradeoffs in terms of performance and failover behavior of these tunable settings.
• Results of experiments to quantify these differences, and how to establish a meaningful baseline across different systems.

The target audience is engineers who are trying to get a deeper understanding of the architectural considerations behind these systems, and also business personnel who understand the idea of NoSQL but want to understand more clearly when to use each kind of system.