epinions
This benchmark is derived from the Epinions.com consumer review website. It uses data collected from a previous study [23] together with additional statistics extracted from the website. This workload is centered on users interacting with other users and writing reviews for various items in the database (e.g., products). It consists of nine different transactions, of which four interact with user records only, four interact with item records only, and one interacts with all of the tables in the database. Users have both an n-to-m relationship with items (i.e., representing user reviews and ratings of items) and an n-to-m relationship with users.
This workload emerged from one of the original “social networking” websites and thus provides an interesting challenge for relational DBMSs. It is similar to the Twitter benchmark, except that its many-to-many relationships are traversed using SQL joins (as opposed to application-side joins), and it has more complex interactions between its tables.
