1

Looking for a method to evaluate best layout for services across AWS regions. My setup is compromised from 2 endpoint, a server in NY4 and client in Asia.

Between them are 2 api hops and a web server. The hops go as follows.

NY4--tcp--API1--json http--API2--json http--Web Server--http/s--Clients.

I could place everything in North America but that will cause bit latency for Asian Clients

I could inversely place everything in Singapore or Seoul but that would increase latency for API1 which is(due to circumstance beyond my control) becomes more unstable the further you get for the the Server at NY4.

One thing that makes sense if to place API1 in north America, API2 in FRankfurt and Web server in Singapore, but I don't know how to evaluate that latency vs the first 2 layouts I described(ignoring unstable API1).

What is the right systematic approach to evaluating a topology like this?

Dmitry
  • 111
  • 2
  • 1
    If you use route53 for your endpoints you can have multiple copies of the various services in each geo region and DNS will pick the one with the lowest latency. – TheFiddlerWins Mar 28 '16 at 16:55
  • Thanks for you input @TheFiddlerWins, but wouldn't that explode the cost of such a layout? The whole point of using AWS(for me) is to minimize the cost of hosting – Dmitry Mar 28 '16 at 17:02
  • 2
    Unfortunately, minimizing cost and minimizing latency across a global deployment will be difficult; optimizing one often means decreasing the other. Which one (low cost or low latency) is more important for you? – Castaglia Mar 28 '16 at 17:37

1 Answers1

1

To expand create an endpoint in each region:

api1.company.com = Route53 latency based DNS for any of api1.nyc.company.com api1.singapore.company.com api1.frankfort.company.com

same for API2 and API3 and the web servers themselves. Each should go where ever is closest.

TheFiddlerWins
  • 2,999
  • 1
  • 15
  • 22