Quote:
Originally Posted by kerowo
I don't have a good handle on our data but I think the cost of moving the data in and out is prohibitive, particularly when we already manage a **** ton of servers in house.
I actually doubt this is true, especially given all the other cost savings you have. Getting data into s3 is almost free and once your data and Hadoop cluster are both in AWS you can run jobs against your data at no cost and at a very minor performance hit (vs having your data in HDFS on your cluster).
Quote:
Originally Posted by kerowo
I'm not sure if there are privacy concerns, but wouldn't be surprised. Regardless, I have very little input into the design decisions of the nodes, my responsibility is getting operational experience to be able to backup the guy who runs the production job stream.
Fair enough, and I assumed that if you have a contract with Cloudera its a big enough company that any sort of significant change like moving data/servers to the cloud would be painful, long, and involve a lot of Management peeps.