There are various criteria and characteristics which are very useful for determining good sample design these are discussed in detailed and with real examples in the following session ->
It depends on what you know about your population, and what you need to learn from it. Are the data skewed? Can you stratify? Do you have auxiliary information? Are you doing inference based on probability or based on a model? This is a very broad question. A different field may have a different approach. I have a number of books and there are many more. You best ask more specific questions. You can also find a lot of good information online by searching on a term and including "Pennsylvania State University" in your search. Their online materials are very good.