Which monitoring tool is better to use on AWS for k8s cluster?
There is no better or worse here, you need both :)
Heapster is a solution that makes it possible for Prometheus to collect metrics for the kube cluster. It is also intended to be replaced by Metrics Server eventually. Heapster will also allow things like kubectl top pods to work, but will not give you a long term metrics storage / analysis / alerting. For that, you need something like Prometheus+Grafana.