java - Dynamic cluster resizing on the cloud -

I am a Java developer and are considering the AWS, Rackspace and Heroes as potential cloud providers for my app . (Virtual Dedicated Server Plan) before working with traditional web hosts only, I am overwhelmed over

I understand that, with the cloud, you have virtual epsels (virtual machines pre-configured software stacks, such as your app server, database server, etc.) based on demand. So if its midnight or vacation and no one is using my app, then I can cut the number of running cases and save some money. But, if on Saturday morning and I am approaching dangerously close to extreme weight, then I can create new servers and have assured the rest that my app is not going in the accident.

In any way is wrong or faulty, please start by correcting me!

Let's say that I am the least right about the clouds, then it is understandable that I should make my own monitor which suddenly increases / decreases in app usage / pattern (I network Based on traffic and bandwidth?!) When traffic increases to a certain point, then I should have more VMS Instruction, and on the contrary when the traffic is slowed down.

Because jacloud allows you to do this, so my monitor only uses JLood API for the egg /

Am I choking: Do the algorithms / formulas give me your monitor To V.M. Should be used to use for egg / kill I wrote a temporary algorithm which basically ensures that I have "HP" (VM) which I need at all times. The monitor chooses about every node every 10 seconds, and if the current load on the system is replaced by a certain delta, then it allows ClusterResizer to kick and determine That is, we need more / fewer nodes. / P>

Although I've spent a few hours preparing the framework / skeleton of this dynamic resizing system, but before I wanted to stop before proceeding too far from myself and if I was doing something I am unnecessary, wrong or just plain nuts!

How do other cloud apps change their VM monitor and size? Should I be familiar with algorithms or best practices? thank you in advanced!

You do not need to implement your own monitor on Amazon EC2, there is a feature that Scaling elastic for you - Automatically scales occur when the load decreases and when it decreases then the bottom is back scales. Therefore you will not need to implement the monitoring service yourself. There is a third-party solution for auto-scaling called Scalar on the Rackspace. There may also be a native rackspace feature of Amazon's CloudWatch, you should check to protect this infrastructure from scratch.