In recent years, the Big Data challenge has attracted increasing attention [1-4]. Compared with traditional data-intensive applications [5-10], these “Big Data” applications tend to be more diverse: they not only need to process the potential large data sets but also need to react to real-time updates of data sets and provide low-latency interactive access to the latest analytic results. A recent study  exemplifies a typical formation of these 4applications: computation/processing will be performed on both newly arrived data and historical data simultaneously and support queries on recent results. Such applications are becoming more and more common; for example, real-time tweets published on Twitter  need to be analyzed in real time for finding users’ community structure , which is needed for recommendation services and target promotions/advertisements. The transactions, ratings, and click streams collected in real time from users of online retailers like Amazon  or eBay  also need to be analyzed in a timely manner to improve the back-end recommendation system for better predictive analysis constantly.
|Title of host publication||High Performance Computing for Big Data|
|Subtitle of host publication||Methodologies and Applications|
|Number of pages||16|
|State||Published - Jan 1 2017|