The coverage at The Register was somewhat better at explaining the microsecond focus: http://www.theregister.co.uk/2014/02/11/google_research_three_things_that_must_be_done_to_save_the_data_center_of_the_future/?page=3
By the way, the abstract for "The Tail at Scale" ( http://research.google.com/pubs/pub40801.html ) indicated 100 milliseconds as the noticeable latency.
From a user experience viewpoint, microsecond latencies in data searches are invisible. Why not add a 1,000 microsecond delay to allow everything to come together before responding to the user? They'll never notice the millisecond delay.
Those who know how to do S2S low latency are not about to explain it here. Most likely busy building and debugging it and worrying if they will be first to market.
Homomorphic encryptions seem a long way from ready to use. They take a lot of processing, and will need infrastructure to manage. When was the last time you used an encrypted signature on your email? And not because it does not work, it is just so tiresome to organize.
Also it should be sobering to read some of the last 30 years of literature about how to use multiple queries you are permitted to make in order to infer the data you are not supposed to know. Especially using aggregation operators, exactly the sort of thing which homomorphic systesm are proud to make possible.
Much of that literature was going around during the late 80s, hacking SQL databases. Probably hard to trace now since it predates the 1994 search wall prior to which no knowledge was in existence. The other day I threw out the fabulous, timely Spectrum issue featuring the new CEO of DEC (a sort of generational deja vu, think about it) Robert Palmer, then regretted it only to discover that IEEE archives vanish that far back. Who knows, I might have destroyed the last copy.
So anyway, encryptions with holes through them seem a little tricky to get right, and not just because of the math. Meanwhile, who are the users?
Your questions make a nice balance. One has obvious, even desperate consumers and a scarcity of providers. In the second, the roles swap.
Replay available now: A handful of emerging network technologies are competing to be the preferred wide-area connection for the Internet of Things. All claim lower costs and power use than cellular but none have wide deployment yet. Listen in as proponents of leading contenders make their case to be the metro or national IoT network of the future. Rick Merritt, EE Times Silicon Valley Bureau Chief, moderators this discussion. Join in and ask his guests questions.