ghost
commented
8 years ago The servers have trivial replication and for any deployment there is going to be many redundant copies of the data, so I think you can already treat the data as ephemeral and not worry so much, even if the data is sitting on a hosted server. I think the simplest thing is to run an osm-p2p node on a cheap commodity hosting provider like digital ocean, ec2, slicehost, or something like that. It is very easy to spin up a db and replicate the data to one of those instances, but it seems much more involved to run a node over dynamodb.
More experimentally, I have been working on some techniques for hosting data without servers: https://github.com/substack/swarmlog and https://github.com/substack/swarmbot
What are the hosting options for running an osm-p2p node in the cloud? Ideally I would like to separate the storage layer from the server, so we can treat the server as ephemeral, and not have to worry about maintaining the storage side. I see that dynamodb is maybe an option using one of these leveldown modules:
https://github.com/jed/dynamo-down https://github.com/davidguttman/dynamodown https://github.com/mapbox/dynamodb-down
I feel a lot safer when the database is looked after by someone else, given our small server-ops team (<1)