Depends on your specific setup and use-case. If you use AWS Load-Balancing you will get a managed service that will autoscale and will only have such a small latency that you can only measure it on a network point of view (milliseconds). The advantage of using AWS Loadbalancing is that you will get detailed metrics that give you a detailed overview about latency, error rate on each of your services behind the reverse proxy. This way you can easily find the bottleneck. Another possibility would be to setup a nginx reverse proxy running on EC2 instance. This way you could enable caching functions for specific locations and your reverse proxy will cache request. This way your reverse proxy can dramatically improve your perfomance but you might need to have rules for auto-scaling if you have a high traffic site...