What do you like best?
With ADLS we can easily define our data modelling from raw sources to refined data. It provides a great interface and a great account management together with IAM management.
Also you can use a client which abstract you from log in to your azure account all the time. Both support several file formats and IMO is the perfect tool to store your output from batch processes
What do you dislike?
Maybe we are not able to inherite r/w permissions to child folders when folder are created dinamically in a process. We had to solve this problem using the azure sdk and its connection to ADLS api.
Recommendations to others considering the product
Please consider this if you wsnt to store your dsta easily and access them from different programms like IntelliJ. If you switch to Azure is a key feature you should include in your cluster.
What business problems are you solving with the product? What benefits have you realized?
We are storing our output from batch processes and we built our dsta modelling there. From this data lake data scientists can make insights and access the data simply with the adl:// url.
We think is easy to mantain and is cheaper than other solutions out there so for the data we manage is correct.