AI infrastructure
1. AI infrastructure
SPEAKER: Building on the previous lesson where you explored Google's three-layered AI architecture, including AI infrastructure, AI development, and AI applications and solutions, you'll now delve into the foundational layer, AI infrastructure. Since its inception in 1998, Google has been dedicated to data and AI. A decade later, in 2008, Google Cloud was introduced to offer secure and flexible cloud computing and storage solutions. You can think of the AI infrastructure in terms of three layers. At the base layer is networking and security, which lays the foundation to support all of Google's infrastructure and applications. On the next layer, sit compute and storage. Google Cloud separates, or decouples, as it's technically called, compute and storage, so they can scale independently based on need. The top layer includes data and AI products, which enable you to perform tasks to ingest, store, process, and deliver business insights, data pipelines and ML models. Thanks to Google Cloud technology, these tasks can be accomplished without needing to manage and scale the underlying infrastructure. However, understanding some essentials about Google Cloud compute and storage can help you grasp the higher level data and AI products. Let's begin with compute. Organizations with growing data needs often require lots of compute power to run data and AI jobs. And as organizations design for the future, the need for compute power only grows. Google offers a range of computing services, from flexible infrastructure to fully managed serverless platforms, balancing control and convenience. For example, Compute Engine-- high control, like managing a physical server; Google Kubernetes Engine, GKE-- control over containerized apps with orchestration benefits; Cloud Run-- serverless convenience. Google manages infrastructure. You might be familiar with the container platform GKE and serverless options like Cloud Run. For more details, check out Google Documentation in the reading list. Where does the processing power come from? It's from the hardware, computer chips. However, traditional computer chips like Central Processing Units, or CPUs, and even the more recent Graphics Processing Units, or GPUs, may no longer scale to adequately reach the rapid demand for AI. To help overcome this challenge, in 2016, Google introduced the Tensor Processing Unit, or TPU. TPUs are Google's customized application-specific chips to accelerate AI workloads. TPUs act as domain-specific hardware, as opposed to general-purpose hardware like CPUs and GPUs. This allows for higher efficiency by tailoring the architecture to meet the computation needs in a domain, such as the matrix multiplication in machine learning. Cloud TPUs, faster and more energy efficient than GPUs and CPUs for AI ML, are integrated across Google products, offering state-of-the-art supercomputing technology to Google Cloud customers. Let's now examine storage. For proper scaling capabilities, compute and storage are decoupled. That is one major difference between Cloud and desktop computing. With cloud computing, compute and storage can scale separately. Most applications need a database and storage solution of some kind. Your best option depends on your data type and business needs. For unstructured data like documents, images, and audio files, cloud storage is your ideal choice. Alternatively, if your data is structured, organized in tables, rows, and columns, you have options like BigQuery, AlloyDB for PostgreSQL, and others. Note that BigQuery, Google's flagship data warehouse, is particularly versatile. It's built for structured data and also highly optimized for semi-structured data like JSON. It can even query unstructured data, such as log files or images stored in cloud storage, by creating an external table that provides a structured reference to that data. This leads to the top layer of the Google Cloud infrastructure, data and AI products. As you explored earlier, Google Cloud offers a comprehensive suite of data and AI tools. How do you piece them together? To build a data-to-AI project, you orchestrate these products through a data-to-AI workflow-- ingest and process, store and analyze, and activate with AI. First, ingest and process data from diverse sources, both real-time and batch, using tools like Pub/Sub, Dataflow, Dataproc, and Cloud Data Fusion. Next, store your data in solutions like Cloud Storage. Then analyze it with various tools. Use BigQuery, AlloyDB, Cloud SQL, and Spanner for SQL databases. Use Bigtable and Firestore for NoSQL databases. Use Looker for visualization. Finally, activate your insights with AI. Train predictive models for forecasting, or leverage Gen AI for content creation and action. Vertex AI is the central AI development platform, offering products like Vertex AI Studio, Agent Builder, AutoML, and notebooks for AI projects ranging from out-of-the-box solutions to custom builds. These tools are seamlessly integrated on Google Cloud, enabling data scientists and AI developers to efficiently transition from data to AI. For example, BigQuery offers embedded SQL commands to train an ML model, a feature you'll explore later. Additionally, within a Vertex AI notebook, you can easily pull data directly from BigQuery using SQL for advanced model training. Don't let the variety of options overwhelm you. You'll focus on BigQuery, the primary data warehouse, and Vertex AI, the AI development platform, later in this course. But before that, let's get you ready with another fundamental topic, AI models, in the next lesson.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.