I have a website (ex: www.website.com) on aws that uses a large amount of streaming data (up to 10mbps / user). Lets say that my website is slowing down because the server simply cannot handle that much data on its current connection (100mbps, 1000mbps.. whatever it is) when multiple users connect.

I've read about load balancers, but it seems like everything still passes through the same front-end, but you are able to split up backend stuff (ex: www.website.com still handles all connections, but now we have 15 app related backend servers to handle some cpu load or something)

My solution right now is to use dns round-robin, which lets me distribute the bandwidth load across multiple servers.

Am i wrong about load balancing and how it works? Would i need to have one front-facing ip address that forwards users to some subdomains (appserver1.website.com - appserver9.website.com) that contains multiple servers running the website?

Sorry if i'm completely wrong here.


  • Well, load-balancer redistributes your traffic between two IPs instead of one IP to two running apps but sharing same data, this is what usually means under term "fault tolerance". If you need yo increase bandwidth you should use not just two IPs, but two apps sharing same data, or even two network cards bonded together on OS level. If this is what you mean it's true – user3417815 May 16 '16 at 22:54
  • I suggest you to close this question, it's too common. You should just read about load balancing and fault tolerance technologies, there is a lot of ways to implement what you want – user3417815 May 16 '16 at 22:55

So DNS load balancing is an option but the issue is if one of the backend servers dies it is still in the mix unless you do health checks. That gets more costly

What you should do is use an ELB in front of your ec2 instances and that is your load balancer. You point the hostname of the endpoint to the ELB hostname given by Amazon.

The ELB will scale up to handle traffic.

  • If the client is one of the mainstream browsers, it handles the detection and failover at no cost with round robin. – symcbean May 16 '16 at 23:00
  • Thanks, its also much easier this way if i need to add more server to the balancer. – Eric May 17 '16 at 17:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.