We're currently developing a new multi-cluster elasticsearch web application.<p>The app will will ever growing hard disk space to store the elasticsearch data, as well as extensive integration to S3 (data backup). More CPU cores are always beneficial and we'll need >= 16GB RAM for each node.<p>We've been considering a few of the usual suspects, including linode and AWS.<p>Assuming "full" server capacity, we'd be looking at something like a Linode 16GB, which has 8 cores and 386GB SSD and costs $160pm (or $1920 per annum). A similar product from AWS would be something like a c3.2xlarge, which would cost around $2750 per annum (1 year heavy utilization reserved instance), which is around 43% more.<p>There are lots of pros and cons to each. AWS provides more flexibility (security groups, elastic IP, etc), but generally provides less bang-for buck. Linode's support is pretty awesome.<p>Is there anyone else in a similar situation that has worked with high-availability, high traffic elasticsearch clusters (or similar)? Who did you choose as your host, and why?
<i>Full Disclosure: Self-promotion ahead</i><p>Jono, my company, qbox.io is a provider of hosted Elasticsearch with deployments on any data center in the AWS, Rackspace, and Softlayer public clouds. You could think of us as a MongoHQ for Elasticsearch.<p>Yes, it will be more expensive than the infrastructure by itself, but it will also come with fast and easy Elasticsearch deployment, scaling, and support from some of the best in the business. I love Linode and AWS, but they will not know how to help you with application-specific questions.<p><i>Haters, commence hating</i>