Ask HN: Fault-tolerant database for small datasets?

3 points·ffggvv··8 comments

What sort of faults and how much tolerance?

What levels of consistency, availability and partition tolerance?

How small is 'small?'

Or to put it another way, what are you trying to do?

What latency is acceptable?

Storing information about users. There won't be many 1000-5000. The data should be replicated on at least two small servers.

All the other questions are still relevant. I mean if the latency is acceptable, Amazon S3 could store the file and is probably more reliable for handling replication as a backup strategy than anything most developers could implement themselves on hardware under their control...actually, I'm thinking a 'cloud' approach is probably the most robust generic strategy and a service like AWS provides good documentation and self-help support and has a rich third party ecosystem.

Pay for what you use pricing can be issue, but at a small scale it's often cheaper than alternatives because usage is low and there's no minimum.

Anyway, absent information facilitates absent specificity.

Amazon's DynamoDB. A small instance of DynamoDB can cost about $10 a month but is fully replicated and managed. It is "scalable" in the sense that a small instance is cheap but you can turn the knobs and pay more for a bigger system that can handle a higher workload.

What about sqlite?

https://sqlite.org/

The problem is that it's an embedded database storing data on a single server. I was thinking about using it and do frequent backups.

Biggest

I'd like to avoid configuration headaches.

Reply on news.ycombinator.com