El combate en jaula de NoSQL: Ronda 2

Architecture matters. A lot.

Thumbtack Technology today reported results from its latest benchmark comparison of MongoDB, Cassandra DataStax, and Couchbase Server. Couchbase crushed the competition: 1.71M operations per second with sub-millisecond latency on read-heavy workloads, compared to just 227K for MongoDB and 99K for Cassandra DataStax. Similar pattern for balanced workloads. You can see more results Aquí, and get the full report Aquí.  

Conclusion: Couchbase is the clear high-performance, general purpose NoSQL leader.

It’s also clear that as the NoSQL market shifts from phase 1 (experimentation) to phase 2 (mission-critical deployments, where performance at scale is key), MongoDB and DataStax are struggling to make the transition. Both are limited by architectural design flaws like MongoDB’s database-wide locking and DataStax’s inability to effectively leverage memory. Those architectural limitations are, in part, what’s behind enterprises’ decisions to replace MongoDB (e.g., at Viber) and Cassandra (e.g., at eBay) with Couchbase Server.  

Couchbase, on the other hand, was engineered from the beginning for performance at scale with modern infrastructures in mind, designed to take full advantage of advances in memory and network technologies.

If you require consistent high performance at scale for your application, or you’re concerned about limitations of MongoDB and DataStax, you should take a look at Couchbase Server.