Distributed systems

System Design question: In a country populated by 100 million voters we have decided to introduce a distributed voting system to conduct a presidental election. We have made some assumption that everyone will vote and that each voter takes 3 minutes to do the voting and that the voting period is during 3 full days. During this period a voter has the ability to both nullify or modify their votes among different number of candidates, arbitrary number of times, until they have submitted their vote. To each voter there exists a unique ID by whom a voter online can check the status of their vote(if the vote has been submitted, and which candidate they have voted on).

Discuss how to build and design such a system and in particular how to take into account Brewer's theorem in terms of Consistency, Availability and Partition Tolerance.