Topic: Globally distributed load balancing - possible alternatives? (Page 1 of 1)

Say company A has a web server located in the data centre of its office in some central european country. Said server might be hosting the web sites for A's numerous branch offices around the world.
Sites might be loading kinda slowly for customers visiting from far away or even fail to load completely as the web server might at one point decide to lock up for a bit and regain its strength only hours later.

Then in response to effects like these, company A might think about introducing some kind of load balancing mechanism - which would make sense for several reasons. So they decide to locate a second server somewhere in one of their data centres in North America.

What proves to be interesting is the question of how the load balancing should actually be handled in such a scenario. The servers should be sharing the load when they are both avaible, they should take over the other server's job whenever one them becomes unreachable, and incoming requests should not be distributed through any kind of random scheme but rather depending on the visitor's location to reduce the latency issues. Some sort of globally distributed load balancing.

Would potential Asylum inmates possibly be happy to share potential reading recommendations, or (just possibly) even some self-authored forum thread literature on such a kind of topic?

no literatur, but a pointer... you could handle the distribution on the dns level - having a 'smart' dns server figure out where the user is and redirect to the appropriate server in question.

this may be an option http://en.wikipedia.org/wiki/Anycast

You might want to look up the Keepalive project, part of the LVS project. We use it for weighted load-balancing of our FTP, web, mail and DNS servers.


Justice 4 Pat Richard