email iconemail phone iconcall
Robin Schumacher, SVP and Chief Product Officer

Updated NoSQL Benchmark of Apache Cassandra, MongoDB, and HBase (and now Couchbase)

By Robin Schumacher, SVP and Chief Product OfficerApril 13, 2015

If you’ve been paying attention to recent press announcements , you’ve no doubt seen a flurry of benchmark press releases from various NoSQL companies. One vendor claims to “dominate” all it’s competitors, while another asserts that it “eclipses” all rivals.

The truth is, the outcome of any performance benchmark depends on a great many factors, and given a certain combination of variables and a particular use case, it’s possible for nearly any database to outrun another. That’s why it’s so important for you to test the database platform you’re considering under the conditions you expect in your production environment, with a dash of extra zeal (i.e. more concurrent threads, slightly higher data loads, etc.) thrown in for safety’s sake.

That said, reviewing benchmark reports can help you in narrowing the candidates that end up on your short list as long as the test you’re reviewing generally aligns  with what you expect to deploy in production.

A couple of years ago End Point, an expert database and ecommerce consulting company, performed a benchmark between Cassandra, MongoDB, and HBase where Cassandra faired quite well. Given that those databases have delivered upgrades since that original report was published  (e.g. MongoDB with Wired Tiger), we asked End Point to repeat the benchmark and also include Couchbase.

For this test, we wanted to make sure that the tests mirrored the vast number of scale-out, production deployments we see with Cassandra, so we made only two requests of End Point:

  • Use data volumes that exceed the RAM capacity on each machine.
  • Ensure that write operations are done where no data loss is possible.

These are common characteristics of nearly all Internet Enterprise applications (e.g. web, mobile, IoT, etc.) that we see. If you have that kind system in the works, we’d like to invite you to give End Point’s new benchmark a read. In it, we think you’ll find valuable  information that will help you compare the performance of the leading NoSQL databases and results we believe speak for themselves.

end point benchmark read write mix 2015.png

  End Point Benchmark excerpt – balanced / read-write test.

The new End Point report titled “Benchmarking Top NoSQL Databases” can be found here.





SHARE THIS PAGE
SUBSCRIBE

Comments

  1. Kevin says:

    For a single node machine can you explain how you obtained the exact opposite results of this result between Cassandra and MongoDB?

    http://info-mongodb-com.s3.amazonaws.com/High%2BPerformance%2BBenchmark%2BWhite%2BPaper_final.pdf

  2. Robin Schumacher, SVP and Chief Product Officer Robin Schumacher says:

    Data Size. When all the data fits into memory and you only need one machine, Mongo could beat Cassandra depending on the test. Plus, if you only need one machine, you likely don’t need Cassandra.

Comments

Your email address will not be published. Required fields are marked *