Rasmus Larsen <rwl@tele.gl> writes:
On Thu, 2011-10-06 at 15:17 +0300, Kostas Zorbadelos wrote:
I don't know where it is best to start a discussion for this but I admit that load balancing interests me.
Now, this article also contains opinions on other matters I am not sure I support. For example, it also criticizes badly the use of DNS for load balancing on the grounds of "DNS was not designed to express policy". And what happens when a single machine is not enough to accomodate load? Do you employ NAT load balancers? Is this a better idea? Having a single "name" for a "service" seems to me like a good idea in general. Anyway, long talk, I guess it needs a thread of its own.
Agreed, in particular NAT load balancers either limit the use of IPSEC in transport mode (which will become viable with ipv6), or will break horribly when people start using it. From my point of view, DNS seems to be the best load balancing strategy out there.
This is a drawback in NAT load balancers I hadn't considered. Nice to know. Generally speaking, load balancing can be addressed either at the DNS or the routing layer. I am not aware of a "universal and clean" solution that would fit all purposes. An extra interesting thing is what will be used in IPv6 environments (which is a question posed also in the v6 WG in regards to RIPE 501)
Regards, Rasmus Larsen
-- Kostas Zorbadelos twitter:@kzorbadelos http://gr.linkedin.com/in/kzorba ---------------------------------------------------------------------------- () www.asciiribbon.org - against HTML e-mail & proprietary attachments /\