А относительно того, что дешевле:
вот интересное сравнение на Реддите:
FYI, I work for SF. Most of time, If it is too good to be true, it usually is. Databricks can basically come up with any price they want to present as they have an endless list of compute, instance & engine type combinations. They can make it look cheap but it doesn't mean it will do the job properly or w/ failures.
The good rule of thumb for DBX cost is to at least double up the DBU costs.
50% DBU + 30% cloud compute + 10% Storage + 10% Networking & Object store file access fees.
After that, if you want proper security, then you need privatelink between compute & storage and pay cloud provider per TB of data moment for every query that reads or writes data.
You have users querying data from onprem or out of region, then egress costs per each query based on size of resultset.
You need auditing on your lakehouse file access and compute nodes? you need to pay cloud provider to keep log files for file access for blob storage and compute nodes(ie. cloudntrail).
You need to analyze and audit all access logs in one place. You need to pay to ingest those log files in different formats to DBX so you can cross join it with DBX logs.
You need DR or redundancy. You call their services team or an SI to build you a custom one where you have to manage it yourself.
When you put all of this together with hard costs & soft costs(admin, maintenance & infosec times) as well as the added responsibility on your shoulders to make sure all of the data files are safe & secure where they cant be accessed by unwanted internal and external users, it suddenly doesn't look that cheap especially these tasks are usually handled by multiple people by multiple teams.
Snowflake, you don't have to deal with any of these things as they are all part of the cost. Software, compute, storage, redundancy over 3 AZs in each region, egress costs, security, encryption, auding is all handled by Snowflake where you are NOT responsible for it.
It is peace of mind you are getting on top of best in class performance and a ton features & functionality to cover all of your data engineering, DataScience, warehousing & lake needs, and much much more.
IMO, data engineers should only focus & be responsible for producing data for business. Any other activity & responsibility they have to own, will take away from their core function which then hurts the business in return.
