EC2 instances backed with EBS can easily recovered by restarting into a different hypervisor InstanceStore in the other hand if the instance fails, the machine and data is lost.
Supports user data if a script execution is needed after the machine is boot.
** clustered placement group (one AZ) and spread placement group (multiple AZs).
Nitro and XEN are the underlying hypervisors for AWS EC2
Options
on-demand: hourly paid.
reserved:
predicatable usage.
Up-front payments with a contract (Standard, Convertible, Scheduled RIs)
from 1 to 3 years terms. The most time the most cheaper
cannot be moved between regions
spot:
low prices
flexible start and end times by biding a price
if the instance is terminated by aws the partial hour of usage is not charged, if you terminate the instance yourself you will be charged for the complete hour
dedicated: regulary stuff, no multitenant virt. Paid in hourly basis and also as a ReservedInstance
Types of EC2
field programmable gate array (F1): financial, big data
high speed storage (I3): nosql dbs, data warehousing
graphics intensive (G3): video encoding, 3d streaming
high disk throughput (H1): mapreduce-based workloads, HDFS or MAPR-FS
lowest cost, general purpose (T2): web servers
dense storage (D2): fileservers, hadoop
memory optimized (R4, X1): mem intensive. SAP HANA
general purpose (M5): app services
compute optimized (C5): cpu intensive
graphics general purpose GPU (P3): ML, Bitcoin mining
Checks
system status checks: Hypervisor issues
instance status checks: OS issues or config, network issues
IAM roles into EC2 instances
An IAM role can be attached to a EC2 on the fly.
AWS Creds are injected in the instance and automatically rotated
For the awscli it is needed to specify the --region $REGION when accessing to a different bucket region from the origin.
AWS EC2 Metadata
http://169.254.169.254/latest/meta-data
/public-ipv4
http://169.254.169.254/latest/user-data
Autoscaling groups
Requires a load balancer in place
Requires a lunch configuration: Like when we create a EC2 (ami, network, vpc, sg, etc).
Provides redundancy on multiple AZs (multiple subnets).
Autoscale policies setup based on Alarms tresholds. Increase and Decrease rules.
Notifications on autoscaling group attaches or removes instances.