Hi Andrew, allow me address these points.
Correct, the RDS instance was in the EU region.
Currently the local DB instance has not been specifically tuned and is using the install defaults. As the DB in question is still v.small these have proved sufficient.
Your points about the AWS/RDS burstable micro instances are interesting and worth exploring. Rather than spend cash idling to clock up some credits I've retested with various higher spec instances to try and establish if there has been a CPU or IOPS bottleneck. I was going to tabulate the results but thus far it's all been pretty flat; the last iteration as was,
- db.m3.large (standard NOT burstable)
- 2 vCPU (6.5 Amazon compute units)
- 7.5GB RAM
- 'Moderate' network performance
- SSD Provisioned IOPS
This is ALL very much higher spec'ed than the original micro instance and the Pi should be no match for it. I have however seen zero change in the general performance experienced with the micro instance with homepage DB query times at ~1.2s. I don't think this is anything to do with the RDS instance and my guess is the main bottleneck is network latency rather than the RDS instance being under provisioned. This is informative regards the original post, as we are mostly concerned with the issues of accessing RDS from a remote data centre.
I'm certain RDS is great, especially when you are using it from with the AWS cloud and the benefits regarding HA and Administration are considerable, however that's not quite the the question being asked here.
As ever, testing your specific use case is the only way to get an idea as to what is right/best for your setup.