Technology-Driven Video Advertising

A global leader in cable and video advertising collects data from more than six billion devices. This data is accurate, rich, and valuable as part of its strategy for providing customized, targeted advertising. The problem was that this data was huge, and the resources required to provide near-real-time targeted advertising were immense. To meet the need, the company established several hundred Hadoop servers and used them to batch process and pre-aggregate the data. Still, with this huge investment in hardware and tools, the process took one to three days to send a targeted ad to a customer.

Solution

The company needed a solution that could handle the huge amounts of data inflowing into its system while also being able to query at near-instantaneous speeds. It contacted several technology providers to find a solution that met the throughput and latency it demanded. In the end, the company implemented a solution that combined the four requirements of real-time analytics. The results were incredible:

• The company went from over 1,000 nodes to only 11.
• Its virtual CPUs were cut by more than half from 3,400 to only 1,400.
• Its data went from being 1–3 days old to only 20 minutes old.
• Its queries dropped from 12 hours to less than a minute.

These cost savings are significant. In addition to the money saved in hardware and processor costs, the company can more accurately identify customer traits and provide targeted advertising in a fraction of the time. It also discovered that the amount of data it could ingest and process jumped significantly. The change from traditional to utilizing the key components of the real-time stack allows the company to accept a wider variety and a larger volume of data with less burden on infrastructure and time.