SportsManias

SportsManias is a sports news aggregation platform that consists of multiple fronts including native iOS and Android mobile apps, an iOS animated emoji keyboard, and a desktop website.

Back in 2011 SportsManias came to us with an idea. That was the start of a long partnership that continues today. We architected and developed the front-end, backend and infrastructure of what has become one of our more gratifying endeavors.

The systems integrates the twitter feed API, Youtube API, as well as 3rd party sports statistics integration, google knowledge base integration all running on a non relational database (MongoDB) and the CakePHP framework. it also pushes hundreds of thousands of push notifications daily, based on the specific teams followed by users. SportsManias currently handles roughly 500,000 users and 3.7 million sessions/ month in the mobile app and over 900,000 users and 1 million session/month on desktop.

Infrastructure Design

Big Data

Due to the huge amount of real time data that needed to be processed, we used Big Data techniques to handle the challenges of data capture, storage, analysis, data curation and querying. We process roughly 600 - 1200 tweets per minute, approximately 1 million per day. Since old news is no news, velocity is essential. The system processes a tweet in 30 - 600 milliseconds, during which time we receive the tweet, assign it to a team, follow its links in order to pull out the final article and/or any videos and images, and finally serve it up to all relevant users.

In order to achieve this we use the latest in Non-relational databases (MongoDB) in combination with the use of Cloud Computing (AWS) for ease of scaling, performance, and flexibility. In addition the application of a Machine Learning classification algorithm (patented by SportsManias) and run through GPU computing, was used to guarantee quality topical content.

Auto Scaling

Scalability is an important topic when designing website infrastructure for a startup. Thanks to the popularization of Cloud Computing over the last decade servers can be spun up in a matter of seconds and potentially setup in minutes. With SportsManias we've designed an auto-scaling architecture on AWS (Amazon Web Services) that increases the number of web servers, push notification servers and adds to the database cluster based on traffic volume among other parameters. This allows for the site to maintain only the minimum necessary server resources in order to handle sudden peaks while not breaking the bank during low traffic periods.

Distributed Processing

In order to reduce system wide latency, processes are queued across different workers and run in parallel.