MOBILUCK.COM: SOCIAL NETWORK FOR GEOLOCATED MOBILE PHONES
In May 2007, I joined the startup MobiLuck.com as Chief Technology Officer (CTO). This startup is focused on a geolocated social network for mobile phones.
Closer to my friends: geolocalization and social networking.
Users are geolocated automatically so that they benefit from advanced features such as finding the nearest Indian restaurant or meeting great people close to them. Users receive SMS alerts when their friends come nearby.
Our technology is carrier-independent, as we have developed an original platform, and the service works worldwide, available in more than 10 languages including Chinese, Arabic, and Hebrew.
Traffic explosion: 1.2M users & 35M pages/month
This LBS friend-finder service has attracted more than 1.2 million users (as of June 2009) since its launch in August 2007.
Currently, the service recruits an average of 2K to 8K users per day according to the current trend.
By the end of Q2 2009, we saw our traffic doubling every 3 months, reaching 35 million pages per month!

Technical environment
Server side: clusterization of the application
The server-side development is based on LAMP: Linux, Apache, MySQL, and PHP.
The service is distributed across a cluster of servers (ranging from Core 2 Duo machines with 4 GB of RAM to octo-core servers with 16 GB of RAM).
Each box is given a dedicated role — and only one whenever possible:
- front server: runs the Web server (Apache 2)
- back server: runs a MySQL instance
In order to handle our traffic explosion, I recently clusterized MySQL itself and switched to Master/Slave replication on two octo-core servers. I also migrated from MySQL 5.0 to MySQL 5.1.
Client-side: iPhone and Facebook
- Client-side development in XHTML and XHTML Mobile
- iPhone development in Objective-C/Cocoa: geolocation API, networking, and communication with the back office
- Facebook app development: enables Facebook users to run MobiLuck from within Facebook
Web services
We interface with many third parties through web services: Twitter, Yahoo!, Google, Maporama, SMS brokers, advertising agencies, cartography, geocoding & reverse geocoding, etc.
Advertising platform and service-wide monitoring
The business model relies on ad distribution based on users’ profiles and locations. We integrate multiple advertising agencies and distribute ads on a per-country or per-profile basis in order to maximize our revenue.
Our system embeds a homegrown monitoring system and statistics tracker that enable us to run the business and make decisions in an agile way.