4 Responses

  1. Itamar Haber
    Itamar Haber June 25, 2014 at 2:49 PM | | Reply

    “…while keeping most of the working set in RAM” – if that is the case, why wasn’t Redis included in the benchmark?

  2. Ben Engber
    Ben Engber June 26, 2014 at 5:33 PM | | Reply

    Great question. This kind of workload is certainly appropriate for Redis. However, we were trying to include NoSQL databases which promise multi-node scalability and fault tolerance. This is one of the main promises of NoSQL databases. It’s possible to shard Redis and use something like twemproxy, but that’s not the same thing as having built in fault-tolerance and automatic scaling.

    Redis Cluster promises to offer more comprehensive support for such things, but seems to still be in beta. As a rule we do not benchmark beta software.

  3. Jack Strohm
    Jack Strohm June 30, 2014 at 4:20 PM | | Reply

    Have you thought about adding Aerospike
    to this benchmark? – http://www.aerospike.com/benchmark/

  4. Ric Johnson
    Ric Johnson July 1, 2014 at 12:26 AM | | Reply

    When will you release the rest of the report?
    We would love to know how you came up with these findings and compare it to other similar NoSQL stores.

Leave a Reply