Case Study: PubMatic achieves lower cost per ad impression and millions in savings with Acceldata

Pubmatic Leverages Acceldata’s Data Observability Platform to Optimize Performance and Cost at Massive Scale

PubMatic, a leading AdTech company handling billions of daily impressions at hyper-scale (3000+ nodes, 150+ PB, 65+ HDP clusters), faced high MTTR, frequent outages and performance bottlenecks that prevented them from improving their key metric—cost per ad impression—and caused constant engineering firefighting and excess infrastructure and software spend. To address this, PubMatic deployed Acceldata’s Data Observability platform, Pulse, beginning in mid‑2020 to gain end‑to‑end visibility across their YARN, Kafka, Spark, HBase and HDP environment.

Acceldata’s Pulse isolated bottlenecks, automated performance improvements and distinguished mandatory from unnecessary data to stabilize pipelines and optimize resources at scale. As a result, PubMatic improved pipeline reliability, eliminated day‑to‑day engineering firefighting, cut OEM support costs, optimized HDFS to reduce block footprint by 30%, consolidated Kafka clusters to save infrastructure, reduced cost per ad impression, and saved millions of dollars in unnecessary software licenses.

PubMatic

Ashwin Prakash

Analytics Engineering Leader Pubmatic Data