Michael Enescu - From Cloud to Fog Computing and IoT | LinuxCon + CloudOpen North America 2014

### Overview In this presentation from LinuxCon + CloudOpen North America 2014, Michael Enescu discusses the transition from centralized cloud computing to fog computing and the Internet of Things (IoT). Highlighting that connected devices already surpassed global human population roughly seven years prior (reaching 6.5 billion), Enescu explains that conservative estimates predict 50 billion connected devices within five years, while others suggest up to 250 billion. The presentation details how physical limitations in network bandwidth relative to compute and storage expansion necessitate processing and analyzing data directly at the edge.

### Key Topics - **IoT Definition & Growth**: Framing IoT as "small sensors + Big Data + action on that data (analytics)," the talk covers its expansion across manufacturing, robotics, energy, smart cities, and healthcare. - **Data Gravity & Network Bottlenecks**: Moore's law (compute doubling every 18 months) and even faster exponential storage growth vastly outpace bandwidth scaling (Jacob Nielsen's observation of ~50% growth every two years). With around two billion billion bytes (2 exabytes) generated daily, raw edge data cannot be continuously backhauled to central clouds. - **Cloud to Fog Architecture**: Rather than shipping all data up to the cloud to store, analyze, and actuate after the fact, fog computing represents a non-trivial, decentralized extension of the cloud where applications move to the data. - **Open Source & Linux Dominance**: Open source, specifically Linux, forms the foundation across the core, access, and gateway layers of modern infrastructure and edge devices.

### Technical Details - **Edge Filtering & Local Analytics**: Real-world operations like commercial aviation demonstrate data volume limits: each jet engine generates about 20 terabytes of data per flight across 25,000 to 28,000 daily flights, making edge-level real-time analysis essential while only filtered or historically necessary data is pushed to the cloud. - **Decentralization Waves**: Computing history cycled from centralized (mainframes) to decentralized (client-server in the 70s/80s), back to centralized (cloud in the 90s/2000s), and now toward massive decentralization driven by fog computing. - **Advanced Deployments**: Enescu highlights extreme edge and high-throughput environments running Linux, including ruggedized Cisco routing and switching hardware, the Keck Observatory in Hawaii (actuators moving 2,000 times per second), and the Large Hadron Collider at CERN (using Caltech-designed OpenFlow networks operating at 1 terabit per second).

### Takeaways - Data gravity dictates that application logic and analytics must move to the edge where data is generated, rather than streaming raw edge data to centralized data centers. - Machine learning and real-time filtering models must execute on edge devices to decide immediate action while optimizing future models. - New IoT-specific service providers will emerge, and nearly all IoT software infrastructure will be open source and Linux-driven due to the credibility and dominance of open development.