To be fair it should be easy to handle this on a t2.nano
TJ Holowaychuk

In my own testing (limited gateway to cached values in redis) a t2.micro is able to get about 200 req/s before having issues (I suspect network throttling for that size as cpu and memory overhead are minimal) Depending on where requests are coming from, the latency is also an issue… judicious use of route53, and multi-zone elb can add a lot of value there.