It all depends on money. If you strictly followed the AWS guidelines they would tell you that you should have a load balancer and at least two m-type instances and each service probably should have its own pair of instances. That way - if one node fails, there will be no downtime. That setup can cost you maybe around $100/month.
But for this kind of project (just for yourself) I would go with just one small (the smallest possible) instance and I would put everything on it. It's good to know that in AWS any instance can fail in any time. It's rare, but when you have thousands of instances, it happens on regular bases. Therefore - the simplest solution is to create your setup with Cloudformation and have a regular backup. Possibly you can create an autoscaling group with just one instance. if it fails, it will get automatically replaced and you can (automatically or manually) recover the data from your backup. For example during the boot it can look for the last backup in s3 and install it (via the script in meta-data).
This is not perfect, but for sure it is acceptable for a hobby project. And it is cheap.
与恶龙缠斗过久,自身亦成为恶龙；凝视深渊过久,深渊将回以凝视…
