Static IPs on EC2
Static IPs aren't enabled by default with EC2, and there is a charge for them, so they really force you to come up with your plan before requesting them.Static IP's are charged $0.01/hour for each hour they are not in use. That amounts to roughly $7.20 / month depending on how many days there are in a particular month. You max out at 5 IPs unless you fill out a form with your justification - much like you'll need to fill out a form to have more than 20 active instances running.
The static IP's are elastic as well. Elastic in that they don't belong to a particular instance - which is good because if you ever have to shut down an instance, you still have the ability to assign your IP to a different instance.
Running on a machine, the static IP is free, so essentially, you get a $0.01 discount on the pricing you were already receiving - be it $0.10/hour, $0.40/hour, or $0.80/hour.
What to do with a static IP?
Static IP's enable a lot of things. You can't run a nameserver without one - which meant that prior to now you couldn't use amazon as your primary hosting provider. Even if you decided to outsource your DNS to somebody like Zone Edit, you would still have the problem of changing the IP within DNS if your web server went down for any particular reason.Remapping vs. Re-engineering
Your static IPs can be dynamically re-mapped to new instances. This means, presumably, that a running host can have it's IP moved from one machine to another - so you can troubleshoot without shutting down. It also means that if you needed to move traffic to a larger instance, you could create the larger instance, wait for it to boot and achieve a ready state, then map your IP to that instance for a quick and easy need-based failover. Alternatively, you could boot a load balancer instance and boot two large instances with your application and have the load balance instance manage the traffic levels for you. In the end, I think Amazon has come up with a great formula that can manage a variety of opportunities.Regions and Availability Zones
Amazon has introduced two new terms in Regions and Availability Zones. A region is in essence a data center. Amazon has plans for releasing region-based services allowing you to choose the location for your instances. I imagine there is an opportunity coming up for anybody interested in building a dynamic and low cost edge caching service. I know I'm already working on a region centric traffic identification utility. If you've got plans, call me! We'll do the wonder-twin thing. Personally, I've got a nice plan for a simple, free, and public edge service and amazon would be an ideal solution.Availability Zones are essentially areas within Amazon's network infrastructure that are designed to have low-latency connectivity within that particular environment, and isolation from failures in other availability zones. You can start up instances in separate availability zones for emergency failover, and within availability zones for local clustering. (although there really isn't a high speed clustering solution at the moment - and I imagine their won't be until the Xen developers can come up with better network resource management solutions).
All of these new terms mean slight changes to the fee structures, but Amazon in all their glory has opted not to implement the new cross-availability zone fees until June so that existing customers don't get shocked or forced into paying new fees before they have a chance to re-align their infrastructure.
Can I tell you how much I'm impressed with Amazon lately?
What visitors have to say about Amazon EC2 now has Static IPs
