Book Image

Serverless Analytics with Amazon Athena

By : Anthony Virtuoso, Mert Turkay Hocanin, Aaron Wishnick
Book Image

Serverless Analytics with Amazon Athena

By: Anthony Virtuoso, Mert Turkay Hocanin, Aaron Wishnick

Overview of this book

Amazon Athena is an interactive query service that makes it easy to analyze data in Amazon S3 using SQL, without needing to manage any infrastructure. This book begins with an overview of the serverless analytics experience offered by Athena and teaches you how to build and tune an S3 Data Lake using Athena, including how to structure your tables using open-source file formats like Parquet. You’ll learn how to build, secure, and connect to a data lake with Athena and Lake Formation. Next, you’ll cover key tasks such as ad hoc data analysis, working with ETL pipelines, monitoring and alerting KPI breaches using CloudWatch Metrics, running customizable connectors with AWS Lambda, and more. Moving on, you’ll work through easy integrations, troubleshooting and tuning common Athena issues, and the most common reasons for query failure. You will also review tips to help diagnose and correct failing queries in your pursuit of operational excellence. Finally, you’ll explore advanced concepts such as Athena Query Federation and Athena ML to generate powerful insights without needing to touch a single server. By the end of this book, you’ll be able to build and use a data lake with Amazon Athena to add data-driven features to your app and perform the kind of ad hoc data analysis that often precedes many of today’s ML modeling exercises.
Table of Contents (20 chapters)
Section 1: Fundamentals Of Amazon Athena
Section 2: Building and Connecting to Your Data Lake
Section 3: Using Amazon Athena
Chapter 11: Operational Excellence – Monitoring, Optimization, and Troubleshooting
Section 4: Advanced Topics

Using built-in ML UDFs

In the previous section, we learned how we can create UDFs using Lambda. In this section, we're going to learn how to use Athena's built-in functionality to create UDFs that delegate down to a ML model. We're not going to delve too deeply into the ML aspects of things, though we will cover some basics just so you know what's happening. If you read Chapter 7, Ad Hoc Analytics, then some of this should be familiar.

Before you get started, note that you may incur some SageMaker charges during this. Particularly for the portion where we are training our models, we don't want to be waiting around forever, so we are leveraging the recommended cost/power instance type of ml.c5.xlarge. Total charges should be no more than a few dollars.

Pre-setup requirements

Before we are ready to head on over to SageMaker, there's a couple of things we need to put in place. First up is our favorite resource, an IAM role. By now, you're probably...