<The End of DNS As We Know It ⁄ >

When you start distributing services (workloads, applications) across multiple locations, a la cloud balancing, and those locations may change on a frequent basis you begin to run into problems with finding those services and scaling the rate of change effectively. DNS was designed to resolve host names, but never expected that the same host name might resolve to one of two, three, or four IP addresses all within the span of five minutes.

If we want to support a rapid rate of change, we'd also need to consider the strain on the existing DNS infrastructure as it would require that propagation rates be decreased such that changes would be discovered as needed rather than 2 or 3 hours (or days) later. That change, however, isn't specific to any particular technology and would affect all resolution requests. That invariably increases traffic and stress on the entire Internet infrastructure.

Too, DNS certainly isn't prepared to deal with the possibility that two different clients might need two completely different addresses for the same host. DNS was not designed to support contextual-based response to a query. It's host name in -> IP address out and - this is what makes it bad for cloud in the future - it's completely anonymous.



este é só um excerto do artigo, para aceder ao artigo completo, clique no link em baixo:
this is just a small excerpt from the article, to access the full article please click in the link below:
http://devcentral.f5.com/weblogs/macvittie/archive/2009/08/28/the-end-...