Heidi: The key tip is to identify the guarantees that you need from the system. Does the system actually require strong consistency for all the requests? Can it tolerate a weaker model? Is it sometimes okay to read stale data? Is it sometimes not okay to read stale data? What kind of scale are we talking about? What kind of network is there between these nodes? How reliable is that network? Do we need to worry about machines possible acting maliciously or arbitrarily? What kind of fault guarantees do you need? Do you actually need to formally verify guarantees that if X failures occur under Y conditions then it would be okay? Or is a weaker idea that it would be up 99.9% of the time ok? These are types of questions we will discuss when you consider what guarantees a system requires.