The computing power of raspberry pi is limited and so is the access speed. I have 2 raspberry pi and was hoping I could somehow connect them together and make them work in parallel to get better and faster performance. My main usage of Ras-Pi is going to be running a local NAS, HTPC and a web-server w/ limited development. All of this to be used by my Macbook, TV and some remote access and backup. I am also planning to stream content through it. I tried to do all of it and the performance was really poor. I am looking for a way to make it faster either using these 2 raspberry pi I have or some cheap upgrades or mods.
Split the tasks, have 1 dedicated to HTPC and hosting the NAS/SMB server; keep this as a media pi. Use the other one as the web server or hosting pi. You can split these tasks however you want. The SMB/NAS pi will obviously be limited by the speed of the USB drive connected to it.
This guy used 64 to create a 'super computer' http://hackaday.com/2012/09/12/64-rasberry-pis-turned-into-a-supercomputer/
Steps on how to do it: http://www.southampton.ac.uk/~sjc/raspberrypi/
It will teach you how to create a 2 node computer. Test it out, looks interesting!
I did that and i am continously using this for my private software experiments(md5 wep wpa distributed bruteforce, distributed raytracing ...) even though PI's are slow it is a great and cheapest way to learn.
Coded my own grid computing library with the iPhone GridMan app (https://itunes.apple.com/pl/app/gridman/id819277978?mt=8)
The grid computing with a stack of raspberries is a great thing to learn for a minimal price, start with two and distribute the tasks. Then reach for another two and compare the diff.
If you think of supercomputing, go with MPI, paraller computing
If you want to run apps on a pile of PIs think of load balancing, cluster (check beowulf), HA, split your configs across NAS, run different services on different nodes, there is no limit in options.
Of course even 10 of PIs wont run as fast as your macbook but that is a different story.