Optimal Scalability and Performance in Programmatic Advertising Platforms

SCHEME: Industrial Fellowships

CALL: 2017

DOMAIN: IS - Information and Communication Technologies

FIRST NAME: Georgios

LAST NAME: Varisteas



HOST INSTITUTION: University of Luxembourg

KEYWORDS: distributed systems, decentralization, real-time bidding, blockchain, hyperledger, fault tolerance

START: 2017-09-01

END: 2019-08-31

WEBSITE: https://www.uni.lu

Submitted Abstract

This research project will investigate optimal solutions for scalability, performance, security, and reliability for global scale distributed systems. All research outcomes will be implemented and evaluated within the context of Programmatic Advertising, specifically real-time bidding for ad auctions (RTB). The project will be under the scientific guidance and supervision of Dr. Radu State of SnT and Prof. Mats Brorsson of KTH. The industrial partner is OLAmobile S.a.r.l., a global mobile advertising company, which is already active in the field of RTB.With RTB advertisers can dynamically buy ad slots published in auctions by ad exchanges. The pervasiveness of ad exchanges recently, has motivated companies like OLAmobile to deploy their RTB platforms at a global scale. Current frameworks suffer from scalability and performance bottlenecks which hinder growth and increase operational costs. This is exacerbated by the tight time window offered for the participation in such an auction. We have identified certain components which can be significantly optimized: i) the banker service which handles an ad campaign’s budget, ii) the bid request processing runtime which can be parallelized with reduced hardware resource requirements, iii) provide transparency of financial transactions, and iv) remove single points of failure while improving fault tolerance. For the first, we propose to leverage business logic for creating selective sharing of data in a new decentralized banker service; thus, data replication will not be a requirement for executing transactions over a budget. For the second, we aim to exploit advances in task parallel programming and data dependency resolution in creating a job placement runtime which is scalable and conservative in hardware utilization. Transparency can be established by the introduction of distributed ledger technology. Modern ledger implementations provide Byzantine Fault Tolerance mechanisms. Finally, single point of failure removal will be a direct outcome of decentralization of the banker service.

This site uses cookies. By continuing to use this site, you agree to the use of cookies for analytics purposes. Find out more in our Privacy Statement