How to solve big data challenges in the ad-tech industry

Table of Contents

  1. Summary
  2. Introduction
  3. Why scale out?
  4. Why in-memory?
  5. The CAP theorem
  6. The one-stop shop
  7. Key takeaways
  8. About Andrew Brust

1. Summary

Advertising technology (ad tech) involves a formidable set of requirements. Operationally, ad-tech tools must work at lightning speeds to offer and complete bids for ad inventory. Meanwhile both historical and real-time data must be analyzed to inform and optimize those operational actions. Conventional operational databases have difficulty scaling to meet these requirements. Analytics databases fall into different categories with their strengths and weaknesses, and they address few if any operational requirements. This can make building ad-tech systems difficult, to say the least.
But a new category of database known as NewSQL is emerging to address multi-genre analytics requirements. Some NewSQL databases even handle operational workloads as well. In this report we investigate ad tech’s requirements and how NewSQL databases address them, and we reveal how one ad-tech firm is using NewSQL databases to improve and simplify its architecture and infrastructure.
Key findings include:

  • Ad tech has exacting requirements that push the analytics envelope. It is heavily reliant on precision and real-time and predictive analytics, with high data volume workloads and demanding SLAs.
  • Scale-out architectures and in-memory technology featured in newer database products work well for ad-tech work and other analytics workloads with high-volume, high-velocity data.
  • Meanwhile sacrificing the relational data model, transactional consistency, and SQL compatibility of conventional databases brings risks, productivity hits, and staffing challenges.
  • NewSQL databases combine facets of different database types such that newer technologies and older standards coexist for maximum benefit.

Full content available to GigaOm Subscribers.

Sign Up For Free