Storm is an obvious choice for high-throughput processing but it's a little out of its territory when you also need to maintain low latency. It doesn't have to be that way, though. We at Forter devised a few components and practices to help us build scalable high-throughput and low-latency storm topologies for our real-time fraud detection system. We'd like to share some of the lessons we learned and our ways of handling the issues. From low-latency join bolts to "escape-routes" for sub-topologies, in this talk we'll reveal all the nitty gritty of how we keep our pipeline processing under one second while maintaining data integrity, robust monitoring and code simplicity.
MP3
MP3
אין תגובות:
הוסף רשומת תגובה