How to benchmark druid and more


I have few questions regarding druid workloads:

1 - how can I benchmark druid for specific type of queries (for example top N) ? should I disable the cache during benchmark tests ?

2 - Does druid can be suitable for backend workloads, for example recommendation system that shows the user top N viewed ads based on some dimensions (for example price range, color, location) ? or for example to show how much views each ad is having on the search result page ? or druid was build for more analytics workload where for example 40-60 analysts executing different queries ?

3 - Is there any public available benchmark of druid vs pinot ?


Hey Igor,

Some thoughts.

  1. Usually people turn off Druid’s query caching by setting “useCache” : false and “populateCache” : false in the query context.

  2. Druid’s designed more for analytics workloads. Although it can be used for what you called ‘backend’ workloads (and people have done that).

  3. Not that I know of.