-1

We have many machines that for security reasons need to communicate through a somewhat static set of IP addresses. ("Somewhat" meaning, manual maintenance once in a while is okay, but new IPs cannot be allocated when new machines come up). I believe something analogous is possible server side using an IP multicast, however searching for terms like "client load balancer," "client multicast," etc. don't give me anything relevant.

Is this possible without creating a bottleneck where all traffic goes through a single machine?

djechlin
  • 95
  • 7
  • possible duplicate of [How do I ensure a REST API does not have a bottleneck when it receives requests?](http://serverfault.com/questions/567718/how-do-i-ensure-a-rest-api-does-not-have-a-bottleneck-when-it-receives-requests) – ceejayoz Jan 16 '14 at 17:58
  • I thought you were evaluating, not building. http://serverfault.com/revisions/567718/1 – ceejayoz Jan 16 '14 at 17:58
  • @ceejayoz that question is completely unrelated. This is a client side problem and that is a server side problem, for one. – djechlin Jan 16 '14 at 18:26

1 Answers1

1

I don't quite follow your question, but it doesn't matter. You have only a couple of choices, of which, I think you're asking about the first one. (And multicast is for server-to-client, not the other way around; think radio.)

First, you control how clients connect to your network. That's normally done via DNS, and most people have one domain name pointing to a single IP address. But you can do more. The phrase you want to search on is "DNS round robin". It's how "big" networks load-balance without introducing a single point of failure.

Second, once the client connects, it's an IP+port that they're connected to. At that point, that machine (usually your router/firewall) would handle load-balancing to machines behind it. Or, you can get more complicated than that, but that's entirely something you deal with internally.

Granger
  • 1,160
  • 2
  • 10
  • 26