Wondering if users of this service can comment.. how has the experience been? Any pros/cons wrt dataproc and emr?
1 star. avoid
Bad bad service
Databricks is now the preferred platform, we are indeed likely dumping. Run run run away.
Azure Hdinsight is one example of good intention and poor execution. The group was created to allow customers Open source solutions in Azure. It plays a key role in getting or keeping enterprise customers to Azure. But the leadership and execution of the group has been horrible. Instead of providing value add on OSS, they are giving a tough experience to customer. No upgrade story, no integration with rest of Azure platforms, heavy cost, poor livesite support are some of the pain.
Please sell the shit outta databricks my msft homies. When we first saw hdinsight it took like over an hour to spin up a cluster. It was hilarious
To be fair, HDinsight was built on top of Windows server which was a disaster in the making. So many issues when my team used. I don’t know how they deal with operational load and compete against normal Hadoop clusters running on Linux. That was several years ago and it might have changed now but people were leaving even that time as they knew it was not going to be successful.
It is horrible. Spark version is old, API is very buggy, hard to scale up cluster without contacting capacity planning. We are dumping. HDINsight in favour of provisioning our own instances on top of AKS