Oh, the things I learned!
The Elastic Cloud service wasn't for us. We use Elasticsearch for storing and analyzing the logs for our various servers/services. We could only store a limited amount of logs (obviously), though much more limited than we would like - your cluster's provisioning are linked on memory and disk space, in order to get more storage you'll also need to bump up the memory. I found I was having to bump to the next tier when my memory usage wasn't even close to maxed out. The price of a production cluster quickly escalated to a point where the cost couldn't be justified. I think this is probably due to our use case - so, your experience my differ.
If you care about uptime, you'll also want(need) to pay for Gold or Platinum support. I liked the convenience of paying through the AWS marketplace - but, the impression I got was that Elastic would prefer you pay them directly.
They also, offer an X-Pack monitoring service, which seems useful, but I didn't use, so I can't comment on it. However, if you chose not to use that service, be prepared to roll your own monitoring solution - otherwise your only way to detect problems pro-actively is via the Elastic Cloud web page.
Perhaps I expected too much from this service, but seeing as it is a pricey option (compared to self hosting), I feel my expectations were justified.
Hi DIY Elasticsearch, Thanks for the review! We recently added more flexibility in our deployments along with changes in pricing: https://www.elastic.co/blog/the-next-generation-elasticsearch-service-hot-warm-clusters-machine-learning-more-hardware-choices-and-new-pricing. Additionally, for logging specific workloads, we introduced hot-warm architecture with built-in index curation that gives you the ability to store recent data on Amazon i3 based instances and gradually move older data to Amazon d2 storage heavy based instances. This type of deployment allows you to optimize storage costs without the hassle of management. Hope you give it a try! Thanks, Elastic Team