Fast graph analysis by efficient CPU utilization

We are working towards improving the performance of big-data analysis. Particularly, our focus is on the analysis of graphs, which represent network structures such as hyperlinks between web pages and personal relationships on social networking services. Analysis of these graphs offers a wide range of knowledge. For example, we can find valuable Web pages, communities of closely related people, and items recommendable for customers. Our novel method lets graph analysis algorithms efficiently utilize multicore CPUs and improves their performance by up to 17 times. Fast graph analysis enables accurate knowledge discovery from large-scale data, a quick response to queries for recommendation or searches for a person, and so on.

Please click the thumbnail image to open the full-size PDF file.