1. Sun Tzu said: The control of a large force is the same principle as the control of a few men: it is merely a question of dividing up their numbers.
2. Fighting with a large army under your command is nowise different from fighting with a small one: it is merely a question of instituting signs and signals.
Energy can be considered another word for efficiency. When it becomes necessary to scale things up, there are two factors which will ensure you lose as little efficiency as possible.
The first factor is how you break the smaller scale model into separate tasks or functions. For example, you may have an application which currently utilizes one or two servers with each of those servers handling several roles. A common model is to have a Web/application server at the front with a database/storage server at the back. As you scale up this solution, you will need to split out the Web service, application (there may be sub-roles here), database and storage roles . As scale increases even more you will most likely end up with many load-balanced Web front end servers; a number of application servers to support the front end and any application sub-roles; and a database back end with query, engine and storage roles residing on separate hardware.
The second factor is how you co-ordinate the communications between the various servers involved. If you started with an optimized, efficient small scale model, then this is simply a matter of mimicking calls that used to be local as network calls to the correct processes on the remote servers. If, however, your small scale design was rife with inefficient calls, that will translate to great losses in efficiency when you try to send those calls across the wire to other servers.
Therefore, it is essential that you take time to maximize efficiency on your smaller scale implementations prior to attempting to scale them up. Make sure to design your application with modules that can be separated later. Pay close attention to the communications which take place between modules that could be split out to different servers later. With your application broken into logical, separate modules which transfer only a minimum amount of data to perform their tasks, scaling up will be much easier.