Half a Billion Hits in a day. This is the stuff of Founder's Dreams (and Ops nightmares)

One in 200 stars has habitable Earth-like planets surrounding it - in the galaxy, half a billion stars have Earth-like planets going around them - that's huge, half a billion. So when we look at the night sky, it makes sense that someone is looking back at us.   Michio Kaku

PacketZoom went about 2X in traffic in 5 days last week ending up with a new record day on Feb 12th. For the first time in PacketZoom's history, we served half a billion requests for the day. The traffic came from 80 countries and hit our servers all across the world. All credit to our sterling engineering/ops team for keeping calm and simply adding servers to get through this spurt without a hitch.

A big reason for easy scaling, across multiple cloud providers and continents, lies in the unique architecture of how mobile apps connect to PacketZoom cloud. From early on, we had decided to take advantage of the fact that our code was running on the client side and was perfectly capable of making intelligent decisions about finding the most appropriate server to go to in real time. This is a clean break from legacy approaches to load-balancing and failover rooted in DNS and server side "reverse proxies".

I have a proposal out for a talk at the Velocity conference this summer on the topic (not yet approved). Here's how the abstract of the talk goes:

"Conventional approaches to service level Fault Tolerance are hand-me-downs from the era when web browser was the canonical client. In the native app era ushered in by mobile, that restriction is lifted. A bare minimum level of intelligence added to the client side allows our team to serve 100's of millions of requests a day globally from bare servers on multiple public clouds at low cost."

So far so good. Looking forward to celebrating the Billion hit day not too far in the future. Watch this space for updates ;-)