I personally have a set of notebooks going back 13 years with all work I've done on the main thread of my research.
Notebooks are essential for experimental work -- it is so easy to get confused as to what you observed!
For CS research, the notebook can be inches of computer printouts with trace data from experiments.
In fact, I even keep trace data output from student theses and journal papers in case I need to see why we drew a conclusion years later.
Also put your ideas in notebooks as they come up.