What are the advantages of Google Compute Engine over Amazon's cloud offering?

After a cursory browse through the documentation, it appears that AWS/EC2 concepts map fairly closely to GCE, in many cases sharing terms and names for system components. One significant difference that will be important to those concerned with implementing high-availability is Google’s networking abstraction which appears to transparently handle cross-region traffic. A single “network”, in GCE parlance, may contain resources in multiple availability zones (AZ) and regions.

AWS regions are almost completely partitioned from each other. Assets created in one region are not visible to another region. Arguably this is a feature that minimizes coupling and therefore presumably minimizes the probability of cascading failures across regions while also increasing security. The tradeoff is, of course, that this introduces numerous headaches both trivial and non-trivial during implementation of any project that requiring region spanning.