Summary
Highlights
Dr. Rajiv Mishra from IIT Patna introduces Edge Computing, its building blocks, architecture, and advantages, especially for IoT. It's presented as an essential component in the course, building upon previous discussions of cloud evolution.
Edge computing enables a truly distributed cloud, contrasting with the highly centralized traditional cloud model. It allows local data processing, aggregation, and querying, reducing reliance on the central cloud and bringing computation closer to data sources.
Edge computing replicates public cloud platform capabilities, moving services closer to data sources. This includes supporting IoT requirements like device management, data ingestion, stream analytics, and machine learning model inferencing, all at the edge.
A key advantage of edge computing is reducing latency by avoiding the round trip to the cloud. It also enhances data sovereignty by keeping sensitive data, such as patient information in healthcare, processed locally at the source, rather than sending it to remote cloud servers.
Edge computing incorporates building blocks similar to cloud services, starting with data ingestion. It handles high-velocity and high-throughput data streams from various sources (batch or real-time) using tools like Kafka, enabling immediate processing close to the data source.
The edge also supports machine-to-machine brokers for device communication and various storage options, including object storage for unstructured data (video, audio) and NoSQL/time series databases for structured data, capabilities previously confined to the cloud.
Edge computing enables real-time stream processing for immediate data transformation and analysis, similar to complex event processing engines. Furthermore, it supports running machine learning models (like TensorFlow Lite, PyTorch) for intelligent processing and predictive analytics at the edge, with model training often still happening in the cloud.
The edge computing architecture is divided into three logical tiers: the data source tier (sensors, databases, social media, etc.), the intelligence tier (where cloud and edge collaborate for machine learning training and inferencing), and the actionable insights tier (for visualizations, dashboards, and automated actions).
The data source tier identifies where data originates. The intelligence tier involves shared responsibility between cloud (for training ML models) and edge (for inferencing). The actionable insights tier focuses on presenting outcomes through visualizations, dashboards, or immediate actions based on the intelligence derived at the edge.
Edge computing fundamentally transforms cloud computing into a truly distributed model. It mimics public cloud capabilities, brings services closer to data sources, reduces latency, and ensures data sovereignty, making it crucial for modern applications like IoT. This distributed ecosystem forms a powerful new computing paradigm.