Title

AWS re:Invent 2022 - Modern applications: An infrastructure perspective (CMP210)

Summary

Yael, an AWS Solution Architect, and Jeremy Atlan from Mobileye discuss optimizing modern applications on AWS.
Modern applications are built from small, loosely coupled, modular, independent components (microservices) that communicate over APIs.
AWS offers over 500 EC2 instance types; selection should be based on workload performance requirements.
Flexibility in instance selection can lead to better compute offerings and cost savings.
AWS EC2 instances can be consumed using launch services and scaling behaviors.
Autoscaling groups are recommended for EC2 AMIs, while Amazon EKS is suitable for containerized applications.
Carpenter is introduced as a solution to manage multiple auto-scaling groups and to use the flexibility of AWS compute services.
AWS Fargate is an option for those who do not need access to underlying instances.
Jeremy from Mobileye presents their use of AWS services, including Amazon SageMaker for machine learning model training and Amazon EKS for complex data applications.
Mobileye utilizes a variety of AWS services to match their diverse application needs, focusing on scalability, resilience, performance, and cost optimization.

Modern applications benefit from being designed with optimization in mind, allowing for easy and effective implementation of optimizations.
The move towards containers, serverless, and event-driven architectures is a strong industry trend.
AWS's broad range of EC2 instance types and processors (Intel, AMD, Graviton, Mac) allows for tailored infrastructure solutions.
Spot instances and autoscaling groups are particularly suited for modern, stateless, and elastic applications.
Carpenter is a key tool for managing EC2 capacity and overcoming the challenges of using multiple auto-scaling groups.
AWS Fargate is ideal for applications that do not require granular control over the infrastructure.
Mobileye's case study illustrates the practical application of AWS services in a high-scale, data-intensive environment.
The importance of choosing the right AWS service and architecture for each specific application is emphasized, as there is no one-size-fits-all solution.
The session highlights the need for attribute-based instance selection, automation, and leveraging AWS tools for efficient infrastructure management.