Lambda architecture-Realtime Machine Learning at Scale
Age and scale are the most challenging ingredients in the recipe for building data-driven products and culture. For this reason Ticketmaster, the market leader in selling tickets at massive scale for over 30 years, is the perfect place to test the limits of the modern data toolkit. If you can make data tools work at Ticketmaster, you can make them work anywhere.
Over many years, Ticketmaster has amassed a large number of in-house data sources. There are tremendous business opportunities if these data can be gathered all in one place. In this talk, I will describe how data are consumed into a data lake and how important pipeline features like idempotency and structure are wrangled to get the most value from these combined sources. During an event “on sale,” traffic at Ticketmaster is similar to a denial of service attack by real people trying to get tickets. The challenge with this level of scale is making data operational from one system to another in near real time. At this scale, by the time data is loaded into a warehouse, the information would be useless for many data-driven products. In this talk, I will describe how streaming data architectures are used to capture the value of these volatile data. Lastly I will use real-world use cases in areas such as personalization, targeting and machine-learning at Ticketmaster to show how anyone can combine relatively mature open source data tools to answer the challenges of building data products at scale.
Parking: Please park at the 7060 Hollywood offices. The parking entrance is located on Sycamore off Hollywood. Take a ticket and we will provide parking validation at arrival.