EC2/AWS Lambda how to get lowest latency

Question

I'm kind of new to AWS and I have a project that requires lowest possible latency to a third party service that hosted on AWS. 
Resovled DNS of it looks like ec*ip_address*.ap-northeast-1.compute.amazonaws.com

So i tested following AWS services for loading target get https://website.com (aka ec*ip_address*.ap-northeast-1.compute.amazonaws.com) :

AWS Lambda ap-northeast-1 ~130  ms

EC2 Instance ap-northeast-1a ~110 ms

EC2 Instance ap-northeast-1c ~110 ms

EC2 Instance ap-northeast-1d ~110 ms

EC2 Instance ap-northeast-1-tpe-1a ~220 ms

EC2 Wavelength ap-northeast-1-wl1-kix-wlz-1 ~120 ms

EC2 Wavelength ap-northeast-1-wl1-nrt-wlz-1 ~80 ms

THIRDPARTY Tokyo vps ~50 ms

AWS CloudShell ap-northeast-1 ~18 ms

The question Is how this is even possible and how do I get AWS CloudShell latencies, because they are actually possible. P.S. Everything was tested on t3.medium and pretty much default settings

Answer

What's missing here is "what is the application" (i.e. is it web-based or something else?) and "where are you measuring the latency from" - because it's going to be variable depending on where the requests come from.

The short answer is "host the application in the region which is closest to your users". The longer answer is that there are many thing you can do to reduce latency but in the end you can't beat the speed of light. One answer might be to host in multiple regions but that increases the level of complexity required in your application.

If your application is web-based, consider using CloudFront. Because there are hundreds of points of presence around the world and it provides caching for HTTP objects.

If your application is not web-based you might consider Global Accelerator which uses the AWS global network to carry traffic from the points of presence to the nearest region where your application is hosted.

EC2/AWS Lambda how to get lowest latency

相关内容