Graph analytics are becoming increasingly popular with several new big-data application domains such as social networks, transportation networks, ad networks, e-commerce, and web search.
However, these graph analytics workloads are seen as quite different from traditional database analytics workloads, largely due to the iterative nature of many graph-analytics computations, and the perceived awkwardness of expressing graph analytics as SQL queries. Interestingly, in many real-world scenarios, the raw graph data is collected and it resides in a relational database in the first place. The data is later extracted and fed to a specialized graph processing system. Given that relational databases are general-purpose data processing systems built with extensability and tuning in mind, how difficult or easy is it to perform graph analytics on relational databases?