EC2

Elastic Compute Cloud (EC2) are virtual machines that you are probably use to in your datacenter. EC2 instances provide full, administrative access to operating system. You are also responsible for all maintenance like upgrading and patching. Nice option for testing is free tier – you can run t2.micro (1CPU, 1 GB of memory) EC2 instance for 750 hours for free.

EC2 allows to use different ordering options:

• On-demand – paid by hour or even second, can be used for testing, compute peaks or simply if you don’t want to commit to reserved resources.
• Reserved – for 1 or 3 years. Reserved capacity can give you discount up to 75%. If you know how much compute you will need that is the right choice.
• Spot instance – gives you possibility to bid for an instance, if your bid is higher than actual price you will get an instance. But be careful, if price rises above your bid your instance will be terminated. Because of that spot instance are appropriate only for some types of workloads.
• Dedicated instance – instance runs on single-tenant hardware. Can be useful when you need dedicated physical hosts for regulatory reasons.
• Dedicated host – it also gives you separate physical hardware but dedicated host provides additional visibility and control over how instances are placed on a physical server, and you can consistently deploy your instances to the same physical server over time. Dedicated hosts give you the same level of compliance visibility as dedicated instances and also add additional benefits in increased network performance and licencing cost opimalization.

Provisioning configuration

During instance provisioning you can specify instance parameters such as:

• Instance type and size – combinations of CPU, memory, storage, and networking capacity. You can choose different types of instances dedicated for different usage: General Purpose, Compute Optimized, Memory Optimized, Storage Optimized and Accelerated Computing. Each instance type comes with variety o sizes, from super small to extra large. To read more see AWS documentation.
• Storage – disk size and type of storage: Elastic Block Storage (permanent block storage) and Instance Store disk (ephemeral disk attached directly to the host, deleted when your machine is terminated or hardware failure). You can’t stop Instance Store backed instance.
• Network and IP type (public, private or Elastic IP). For more information about networks see my previous post.
• Placement group – if selected AWS will try to place grouped instances physically as close to each other as possible (no guaranties thought).
• IAM role – more about IAM here.
• User data – your scripts launched during creation
• Amazon Machine Image

AMI

Amazon Machine Image (AMI) is a template from which you can launch EC2 instance. It contains root storage volume with operating system (and sometimes applications), block device mappings for additional disks and launch permissions that controls which AWS accounts can use the AMI to launch instances.

AMI can come from different sources:

• AWS – approved AMIs optimized to run in AWS.
• Custom – to create a custom AMI you need to launch an instance from an existing AMI, customize it, and then save it as a custom AMI.
• Marketplace – paid AMIs from other vendors like SAP, Microsoft, Barracuda,etc or many open source offerings.
• Community – AMIs shared by other AWS users, Amazon doesn’t take any responsibility for that content.

Reference Materials

AWS on the road
AWS EC2 documentation
AWS AMI documentation
Dedicated hosts vs Dedicated instances
Standard vs. Convertible Offering Classes