could be connection overhead since the main nodes are in the US, but 5s is still a lot; its also possible that the lambda simply needs to load a lot of support libs to make that connection; can you see if a subsequent one is faster?