Home/ Questions/Q 8785215
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-13T21:17:49+00:00

I have a client application that connects to a custom server running on EC2

  • 0

I have a client application that connects to a custom server running on EC2 using TCP sockets. The connections between the client and server can last for awhile (multiple hours to a full day).

I am now in the process of evaluating Amazon’s ELB (Elastic Load Balancer), but am worried that due to the dynamically changing IP of the ELB, existing connections will be lost when the IP changes. Can anyone confirm if this is how it works? Is it simply the IP associated with the DNS name that changes? Is that the same thing as the IP of an existing socket connection changing? And ultimately, will the effect of such a change be to sever existing TCP-based socket connections?

EXTRA INFO:

This is not a “standard” http or browser based app – the client app runs in realtime on mobile devices and communicates with the server via a dedicated TCP socket. The server performs continuous logic and communicates back and forth with the client over the lifetime of the connection over this single TCP socket connection. I want to ensure that the client won’t have to re-connect to the server (i.e., destroy/recreate socket connection) when the ELB decides to change IPs (it’s one thing if you legitimately leave wifi/cellular range or something, but I don’t want to disconnect just due to the whims of ELB).

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    No. ELB will load balance new incoming connections, but will not tear down and redirect existing questions.

    However, I would question the benefit of using ELB with such system exactly because of this. ELB can’t guarantee fair load on all your instances. Consider the following scenario with two instances:

    1. ELB directs first connection to instance A.
    2. ELB directs second connection to instance B.
    3. ELB directs third connection to instance A.
    4. Instance B connection goes down as the client is done.

    Now you have “heavy” load on A, and no load on B at all. At this point ELB will send new connections to B, to even out the load. And when the A connections go down, you get into the opposite situation – “heavy” load on B, and no load on A.

    ELB could be quite useful when you have multiple instances that serve short requests. For long-term connections, I would use static IPs and not hide those instances behind ELB.

    Note that it might be possible to actually mix both approaches, and have the same instance respond to both ELB-fronted requests and static-IP requests by using two separate domain names pointing to ELB and the static IP.

    • 0