In recent years, microblogging services such as Twitter and Yammer have reached phenomenal levels of success and become a significant new form of Internet communication utility. Twitter, the most successful microblogging service, has more than 100 million users and generates more than 65 million "tweets" per day. However, current microblogging systems severely suffer from performance bottlenecks, central points of failure, and malicious attacks, not to mention the high operational and maintenance overhead. Thus, current microblogging systems may threaten the scalability, reliability, as well as availability of the offered services.

The Cuckoo project consists of three aspects of microblogging services. Firstly, we propose a novel system architecture tailored for microblogging services to address the scalability issues, which eliminates main server burden and achieves scalable microcontent delivery by decoupling the dual functionality components of microblogging (i.e., a social network and a news media infrastructure). Secondly, we study the human behavior of online microblogging services. We are investigating human behavior, e.g., burstyness and correlations of microblogging as well as spreading laws of micronews, aiming at a further understanding of microblogging services. Thirdly, we design novel methods to realize unbiased sampling of microblogging's directed social graph which is different from traditional bidirectional social graphs (e.g., Facebook).


September 2011
Cuckoo system at Middleware 2011
Poster on OSN partitioning at Middleware 2011

June 2011
Graph sampling at SIMPLEX 2011

August 2010
Demo of Cuckoo system at SIGCOMM 2010
Poster on graph sampling at SIGCOMM 2010

May 2010
Position paper of Cuckoo system at HotPlanet 2010