John Allspaw provides a glimpse into how other fields handle incident response, including active steps companies can take to support engineers in those uncertain and ambiguous scenarios.
Kenji Rikitake discusses using Erlang/OTP for IoT, covering communication protocols, design principles and overcoming hardware limitations for endpoint devices in fault-tolerant systems.
Pushpraj Shukla discusses how Microsoft Bing predicts the future based on aggregate human behavior using one of the largest scale data sets, and recent progress in large scale deep learnt models.
David Riddoch talks about the technologies that make high performance networking possible on commodity servers, with a special focus on direct access to the network adapter by bypassing the kernel.
Peter Bakas presents in detail how Netflix has used Kafka, Samza, Docker, and Linux to implement a multi-tenant pipeline processing 700B events/day in the Amazon AWS cloud.
David Talby demos using Python libraries to build a ML model for fraud detection, scaling it up to billions of events using Spark, and what it took to make the system perform and ready for production.
Sid Anand discusses how Agari is applying big data best practices to the problem of securing its customers from email-born threats, presenting a system that leverages big data in the cloud.
Irad Ben-Gal discusses Big Data analytics misconceptions, presenting a technology predicting consumer behavior patterns that can be translated into wins, revenue gains, and localized assortments.
Jan Neumann presents how Comcast uses machine learning and big data processing to facilitate search for users, for capacity planning, and predictive caching.
Axel Fontaine looks at what Immutable Infrastructure is and how it affects scaling, logging, sessions, configuration, service discovery and more.
Mathieu Bastian explores the mechanics of unit, integration, data and performance testing for large, complex data workflows, along with the tools for Hadoop, Pig and Spark.
Jon Moore talks about distributed monotonic clocks (DMC) whose timestamps can reflect causality but which have a component that stays close to wall clock time.