BigQuery's on-demand model charges just for the resources consumed during the job execution (via a per-TB proxy), rather than resources provisioned. This is the highest order of cloud-native pricing models, and good on Athena for doing the same!
If one were to equate this to VM pricing, you're getting:
- The ability to scale from 0 to thousands of cores in roughly 1 second (what it takes for Dremel to put together the execution plan)
- Pipelined execution (meaning, each processing stage ONLY uses the resources required for that stage, rather than net resources. think of this as per-second autoscaling from query stage to query stage. This is part of the reason why BigQuery does very well in concurrency tests)
- Equivalent of per-second billing
This in practice equates to 100% resource efficiency - never pay for resources you just deploy and not use.
I've written on this topic at [0] and at [1] and happy to chat further!
(And BigQuery did introduce a Flat Rate pricing model last Fall for the more enterprisey crowd)
BigQuery's on-demand model charges just for the resources consumed during the job execution (via a per-TB proxy), rather than resources provisioned. This is the highest order of cloud-native pricing models, and good on Athena for doing the same!
If one were to equate this to VM pricing, you're getting:
- The ability to scale from 0 to thousands of cores in roughly 1 second (what it takes for Dremel to put together the execution plan)
- Pipelined execution (meaning, each processing stage ONLY uses the resources required for that stage, rather than net resources. think of this as per-second autoscaling from query stage to query stage. This is part of the reason why BigQuery does very well in concurrency tests)
- Equivalent of per-second billing
This in practice equates to 100% resource efficiency - never pay for resources you just deploy and not use.
I've written on this topic at [0] and at [1] and happy to chat further!
(And BigQuery did introduce a Flat Rate pricing model last Fall for the more enterprisey crowd)
[0] https://cloud.google.com/blog/big-data/2016/02/understanding...
[1] https://cloud.google.com/blog/big-data/2016/02/visualizing-t...