Case Study: Leading Omnichannel Ad Publishing and Bidding Platform achieves near-real-time analytics and reduces query times from hours to minutes with SQream DB

SQream DB Serves as a Speed Layer for Bid Optimizations

Leading Omnichannel Ad Publishing and Bidding, a global omnichannel ad publishing and bidding platform handling trillions of impressions per month, struggled to analyze massive impression logs on a distributed Hadoop/HBase/Phoenix stack—queries could take up to five hours, preventing reliable ad‑hoc analysis and near‑real‑time yield optimization. To solve this, they turned to SQream and its GPU database product, SQream DB.

SQream deployed SQream DB as a GPU‑based speed layer to offload heavy analytics, ingesting 85 TB/day and allowing simultaneous queries on fresh and historical data. Queries dropped from hours to minutes, enabling deeper near‑real‑time insights and more targeted ads, supporting 12 concurrent users and 40 TB per chassis (two 2U servers + one NVIDIA Tesla), eliminating Hadoop expansion needs and lowering costs; the customer later expanded the SQream deployment to handle 500 TB.