Comment by sandGorgon

9 years ago

did you consider using EMR and Hbase? its not very expensive to run a provisioned 3 node managed cluster.

I'm fairly sure that EMR and HBase would give us miserably bad performance characteristics for the kinds of workload we're describing here.

Why would I consider it for sub-millisecond latency tasks? Has something changed there I'm unaware of?

  • well i was just curious. im kinda wondering if anything among hbase, Cassandra, Mongodb etc are equivalent in performance at your scale. i see that you mentioned rethinkdb below.. but after their collapse, im wondering what are the options you are looking at.

    • No comment on the corporate aspect. On a personal level, I avoid using MongoDB on new projects.