Clusters, clusters

* So much to look into regarding hardware, especially what exactly to buy. Still have ideas about how we're going to set up our servers: my wish would be to create two clusters of servers, one for running all necessary applications and one for databases, and we could just add or subtract servers depending on load trends. One thing that wouldn't change though is that once customers start signing up to the service, the databases are going to get bigger and bigger, that one's for sure. How we will need to scale up the application server area to maintain an equivalent quality of service (ie. the user doesn't notice any slowing down in the connection, etc.), that I do not know either right now.
* I am basically looking forward so that I don't have to reconfigure the whole server setup time and time again (basically rebooting parts of the server area) when we add on new servers in the near future. I want to have a "cluster" up and running, and when we need more, we just "plug" a new server into the cluster and the cluster then pools (or virtualizes) the new add-on of resources and spreads the workload around. Agreed, it's probably more complex than this, but that's the idea of centralizing the resources to all perform the same purpose and if one crashes, another mirrored machine can take its place.
* Hence, I would need maybe 2 application servers processing the code off application server disks, and 2 database servers processing the requests from the application area. If I can set this up in one single cluster that would be great. Am looking at some people in the US called Penguin Computing, and their helpline allows you to choose "Asia", but that's pretty broad and could be Jakarta as well as Vladivostok, so need to find out for sure if it includes Tokyo.
* Right now with the IBM setup we were looking at originally at around 1.1m yen on the hardware/software setup, so if it's comparable and available asap in Japan I'll go for it. But the more I look at it, the more I see that people are talking about "clusters" in the computational sense of the word (ie combined processor power), and not clusters being both processors and storage disks. I am understanding that the storage part is being left to SANs (Storage Area Networks) that act as the database cluster, for example? Still haven't nailed this thing down. But it'll probably be expensive so probably keep the setup as is, and when need be switch the current servers to act as clusters and integrate the additional components. The more I read, the more it looks like it's just a lot of (expensive) hassle for nothing right now.
* More after a few hours' sleep...