Amazon Quick Start AMIs Amazon Quick Start images appear at the top of the list in the console when you start the process of launching a new instance. The Quick Start AMIs are popular choices and include various releases of Linux or Windows Server OSs and some specialty images for performing common operations (like deep learning and database). These AMIs are up‐to‐date and officially supported.

       AWS Marketplace AMIs AMIs from the AWS Marketplace are official, production‐ready images provided and supported by industry vendors like SAP and Cisco.

       Community AMIs More than 100,000 images are available as community AMIs. Many of these images are AMIs created and maintained by independent vendors and are usually built to meet a specific need. This is a good catalog to search if you're planning an application built on a custom combination of software resources.

       Private AMIs You can also store images created from your own instance deployments as private AMIs. Why would you want to do that? You might, for instance, want the ability to scale up the number of instances you've got running to meet growing demand. Having a reliable, tested, and patched instance image as an AMI makes incorporating autoscaling easy. You can also share images as AMIs or import VMs from your local infrastructure (by way of AWS S3) using the AWS VM Import/Export tool.

       General Purpose The General Purpose family includes T3, T2, M5, and M4 types, which all aim to provide a balance of compute, memory, and network resources. T2 types, for instance, range from the t2.nano with one virtual CPU (vCPU0) and half a gigabyte of memory all the way up to the t2.2xlarge with its eight vCPUs and 32 GB of memory. Because it's eligible as part of the Free Tier, the t2.micro is often a good choice for experimenting. But there's nothing stopping you from using it for light‐use websites and various development‐related services. T2s are burstable, which means you can accumulate CPU credits when your instance is underutilized that can be applied during high‐demand periods in the form of higher CPU performance.M5 and M4 instances are recommended for many small and midsized data‐centric operations. Unlike T2, which requires EBS virtual volumes for storage, some M* instances come with their own instance storage drives that are actually physically attached to the host server. M5 types range from m5.large (2 vCPUs and 8 GB of memory) to the monstrous m5d.metal (96 vCPUs and 384 GB of memory).

       Compute Optimized For more demanding web servers and high‐end machine learning workloads, you'll choose from the Compute Optimized family that includes C5 and C4 types. C5 machines—currently available from the c5.large to the c5d.24xlarge—give you as much as 3.5 GHz of processor speed and strong network bandwidth.

       Memory Optimized Memory Optimized instances work well for intensive database, data analysis, and caching operations. The X1e, X1, and R4 types are available with as much as 3.9 terabytes of dynamic random‐access memory (DRAM)‐based memory and low‐latency solid‐state drive (SSD) storage volumes attached.

       Accelerated Computing You can achieve higher‐performing general‐purpose graphics processing unit (GPGPU) performance from the P3, P2, G3, and F1 types within the Accelerated Computing group. These instances make use of various generations of high‐end NVIDIA GPUs or, in the case of the F1 instances, an Xilinx Virtex UltraScale+ field‐programmable gate array (FPGA—if you don't know what that is, then you probably don't need it). Accelerated Computing instances are recommended for demanding workloads such as 3D visualizations and rendering, financial analysis, and computational fluid dynamics.

       Storage Optimized The H1, I3, and D2 types currently make up the Storage Optimized family that have large, low‐latency instance storage volumes (in the case of I3en, up to 60 TB of slower hard disk drive [HDD] storage). These instances work well with distributed filesystems and heavyweight data processing applications.