D-thinker is More Than 60x Faster Than Spark for PageRank Graph Computation

On Sep. 12, 2014, Data Thinker community publishes the performance results of PageRank computation on Data Thinker and Spark. The results show that PageRank computation on D-thinker is more than 60 times faster than GraphX on Spark.

The PageRank computation is on a 4.8 million node social graph of LiveJournal. The computation is conducted on 3 physical servers for both D-thinker and Spark.

The implementation of PageRank from GraphX/Spark took 876 seconds to finish the computation while the PageRank program on D-thinker tooks only 13.2 seconds which is 66 times faster.