clusteroops

Not quite a easter egg, but one of my favorite esoteric features is the unit disambiguation in the calculator. For instance, a "pound" could be mass, force, or currency, so you can ask for something ridiculous like:

https://www.google.com/search?q=10+pounds+pounds+pounds+in+kilograms+newtons+dollars

We get attacked all the time, although most are so small or simple that our systems automatically diffuse or block them. For the larger products (e.g. Search, GMail), we only really need to manually intervene a handful of times per year.

Home page outages almost never affect all users simultaneously. There many different systems involved in simply connecting users to Google, and most incidents happen outside of our network. We do occasionally have network outages, which are regional, e.g. a few states or countries. We also occasionally introduce language-specific bugs, e.g. garbling CJK. As far as I can recall, the last global outage was back in 2005:

http://en.wikinews.org/wiki/Google_suffers_DNS_outage

Pings are cheap, and the aggregate rate is steady, modulo DoS attacks. So we don't worry about it. (Side note: I actually wrote the code in our load balancer that crafts ICMP responses.)

Lots and lots of tools, most of which we write ourselves. Probably too many to fit in this box.

One big one is a monitoring system called "Borgmon". It collects, aggregates, and records instrumentation data, draws graphs, sends alerts, and dispenses candy. It's extremely powerful, but painful to use, because it has its own programming language. So we have a love-hate relationship with it.

Sleep schedule: roughly 12-2 to 8-10. Oncall doesn't affect it much, since teammates in different time zones cover nights. And I'm only oncall roughly one week in six.

Team size: roughly 40 in my group, which takes care of Google Search. We're spread across Mountain View, Dublin, and Zurich. But I regularly work closely with many other groups.

Larry and Sergey: both several times. I even briefly shared an office with Sergey.