Not only Structured Query Language Method of Ad Request Processing

• Main Authors: Nikolaiev, V.А., Konashevych, O.I.
• Format: Article
• Language: English
• Published: Інститут проблем моделювання в енергетиці ім. Г.Є. Пухова НАН України 2017
• Series: Электронное моделирование
• Subjects: Применение методов и средств моделирования
• Online Access: http://dspace.nbuv.gov.ua/handle/123456789/115859

VertaMedia Company’s server provides operation of advertising exchange system between publishers (site’s owners), advertisers and intermediaries (SSP¹ and DSP² platforms). The objective of the system server is to process a request from a Publisher’s site as quickly as possible, choosing the most relevant advertisement campaign, to show it to a site’s user. The system works real-time online and as faster it makes accurate choice, the more likely that a user will see an advertisement. The obvious solution was to use relational database to compare the parameters of queries with the parameters and settings of ad campaigns, stored in this database. However, it turned out to be unsuitable as such system showed high latency. VertaMediaTM programmers have found an original way to process data, in which comparison occurred in a flat table using the hash sum and a binary tree for matching ad campaigns and another part of the request, which contained a set of keywords/tags was processed by Sphinx Search as local software solution. A method incorporates the original decision to work with database management systemless non-relational tables and use of specialized software solutions for matching keywords. It showed remarkable results in performance of a resource-intensive process, as described in detail in the article.