It is my understanding that 8 is expected to be a good number as long as most are not suspected of being linked. I understand, however, that depending on your question, your organism, and your available sequences, that you may be able to learn something with fewer than 8 sequences eg. 4 or 6. I really think 2 is too few and many more than 8 might be unnecessarily computationally heavy and provide you with little or no additional information.

http://people.sc.fsu.edu/~pbeerli/Beerli_Lab/Peter.html

http://evolution.genetics.washington.edu/

It is hard to give a direct answer to your question. As always, the more markers the better the resolution. As always, budgets are finite.

You should plan your research question first, and read about your model species. Are there studies on this species using microsats? If yes, you should look how diversethese markers were.

The "power" of your analyses will depend on the number of markers, but also on the information you can extract from them. If you use 12 markers with only a 2-3 alleles each, that may not be enough to properly differentiate between your genetic clusters. Sometimes, 6 markers that are highly informative are enough.

What exactly do you want to look at?

Thank you very much for your replies. I am working now with Oenanthe aquatica and Lycopus europaeus. For these species I have 10 markers following Favre-Bac et al 2014. But I will start with Iris pseudocarus soon and I don't know how to design microsats for this species. Lab protocols would be useful! Thank you Michael :)