Everybody agrees that IoT is changing the world... and creates new challenges for software developers, architects and DevOps. How can we build efficient and highly scalable distributed applications using open-source technologies? What are characteristics of data generated by IoT devices and how it differs from traditional enterprise or Big Data problems? Which architectural patterns are beneficial for IoT use cases and why some trusted methods eventually turn out to be “anti-patterns”? This talk will show how to combine best-of-breed open-source technologies, like Apache Spark, Mesos and Riak, to build scalable IoT pipelines to ingest, store and analyze huge amounts of data, while keeping operational complexity and costs under control. We will discuss cons and pros of using relational, NoSQL and object storage products for storing and archiving IoT data and make a case for Time Series database deserving a separate category in NoSQL classification.