Elastic Compute Cloud (EC2)

Overview

• EC2 is one of the most popular of AWS’ offering
• It mainly consists in the capability of :
  • Renting virtual machines (EC2)
  • Storing data on virtual drives (EBS)
  • Distributing load across machines (ELB)
  • Scaling the services using an auto-scaling group (ASG)
• Knowing EC2 is fundamental to understand how the Cloud works

EC2 sizing & configuration options

• Operating System (OS): Linux, Windows or Mac OS
• How much compute power & cores (CPU)
• How much random-access memory (RAM)
• How much storage space:
  • Network-attached (EBS & EFS)
  • hardware (EC2 Instance Store)
• Network card: speed of the card, Public IP address
• Firewall rules: security group
• Bootstrap script (configure at first launch): EC2 User Data

EC2 User Data

• It is possible to bootstrap our instances using an EC2 User data script.
• bootstrapping means launching commands when a machine starts
• That script is only run once at the instance first start
• EC2 user data is used to automate boot tasks such as:
  • Installing updates
  • Installing software
  • Downloading common files from the internet
  • Anything you can think of
• The EC2 User Data Script runs with the root user

Example:

#!/bin/bash
# Use this for your user data (script from top to bottom)
# install httpd (Linux 2 version)
yum update -y
yum install -y httpd
systemctl start httpd
systemctl enable httpd
echo "<h1>Hello World from $(hostname -f)</h1>" > /var/www/html/index.html

EC2 Instance Types

You can use different types of EC2 instances that are optimised for different use cases (opens new window)

Naming convention AWS has the following naming convention: m5.2xlarge

• m: instance class
• 5: generation
• 2xlarge: size within the instance class

General Purpose

• Great for a diversity of workloads such as web servers or code repositories
• Balance between:
  • Compute
  • Memory
  • Networking

Compute optimized

Great for compute-intensive tasks that require high performance processors:

• Batch processing workloads
• Media transcoding
• High performance web servers
• High performance computing (HPC)
• Scientific modeling & machine learning
• Dedicated gaming servers

Memory optimized

Fast performance for workloads that process large data sets in memory

Use cases:

• High performance, relational/non-relational databases
• Distributed web scale cache stores
• In-memory databases optimized for BI (business intelligence)
• Applications performing real-time processing of big unstructured data

Storage optimized

Great for storage-intensive tasks that require high, sequential read and write access to large data sets on local storage

Use cases:

• High frequency online transaction processing (OLTP) systems
• Relational & NoSQL databases
• Cache for in-memory databases (for example, Redis)
• Data warehousing applications
• Distributed file systems

Instace types examples

Instance	vCPU	Mem (GiB)	Storage	Network Performance	EBS Bandwidth (Mbps)
t2.micro	1	1	EBS-Only	Low to Moderate
t2.xlarge	4	16	EBS-Only	Moderate
c5d.4xlarge	16	32	1 x 400 NVMe SSD	Up to 10 Gbps	4,750
r5.16xlarge	64	512	EBS Only	20 Gbps	13,600
m5.8xlarge	32	128	EBS Only	10 Gbps	6,800

Full of list ec2 instances.vantage.sh (opens new window)

Purchasing Options

• On-Demand Instances – short workload, predictable pricing, pay by second
• Reserved (1 & 3 years)
  • Reserved Instances – long workloads
  • Convertible Reserved Instances – long workloads with flexible instances
• Savings Plans (1 & 3 years) – commitment to an amount of usage, long workload
• Spot Instances – short workloads, cheap, can lose instances (less reliable)
• Dedicated Hosts – book an entire physical server, control instance placement
• Dedicated Instances – no other customers will share your hardware
• Capacity Reservations – reserve capacity in a specific AZ for any duration

On Demand

• Pay for what you use:
  • Linux or Windows - billing per second, after the first minute
  • All other operating systems - billing per hour
• Has the highest cost but no upfront payment
• No long-term commitment
• Recommended for short-term and un-interrupted workloads, where you can't predict how the application will behave

Reserved Instances

• Up to 72% discount compared to On-demand
• You reserve a specific instance attributes (Instance Type, Region, Tenancy, OS)
• Reservation Period – 1 year (+discount) or 3 years (+++discount)
• Payment Options – No Upfront (+), Partial Upfront (++), All Upfront (+++)
• Reserved Instance’s Scope – Regional or Zonal (reserve capacity in an AZ)
• Recommended for steady-state usage applications (think database)
• You can buy and sell in the Reserved Instance Marketplace
• Convertible Reserved Instance
  • Can change the EC2 instance type, instance family, OS, scope and tenancy
  • Up to 66% discount

Savings Plans

• Get a discount based on long-term usage (up to 72% - same as RIs)
• Commit to a certain type of usage ($10/hour for 1 or 3 years)
• Usage beyond EC2 Savings Plans is billed at the On-Demand price
• Locked to a specific instance family & AWS region (e.g., M5 in us-east-1)
• Flexible across:
  • Instance Size (e.g., m5.xlarge, m5.2xlarge)
  • OS (e.g., Linux, Windows)
  • Tenancy (Host, Dedicated, Default)

Spot Instances

• Can get a discount of up to 90% compared to On-demand
• Instances that you can “lose” at any point of time if your max price is less than the current spot price
• The MOST cost-efficient instances in AWS
• Useful for workloads that are resilient to failure
  • Batch jobs
  • Data analysis
  • Image processing
  • Any distributed workloads
  • Workloads with a flexible start and end time (resilient to failures)
• Not suitable for critical jobs or databases

Spot Instances Requests

• Define max spot price and get the instance while current spot price < max
  • The hourly spot price varies based on offer and capacity
  • If the current spot price > your max price you can choose to stop or terminate your instance with a 2 minutes grace period.
• Other strategy: Spot Block (NO SUPPORT ANYMORE)
  • “block” spot instance during a specified time frame (1 to 6 hours) without interruptions
  • In rare situations, the instance may be reclaimed

Detail link (opens new window)

Terminate Sspot instances

• You can only cancel Spot Instance requests that are open, active, or disabled.
• Cancelling a Spot Request does not terminate instances
• You must first cancel a Spot Request, and then terminate the associated Spot Instances

Spot fleets

• Spot Fleets = set of Spot Instances + (optional) On-Demand Instances
• The Spot Fleet will try to meet the target capacity with price constraints
  • Define possible launch pools: instance type (m5.large), OS, Availability Zone
  • Can have multiple launch pools, so that the fleet can choose
  • Spot Fleet stops launching instances when reaching capacity or max cost
• Strategies to allocate Spot Instances:
  • lowestPrice: from the pool with the lowest price (cost optimization, short workload)
  • diversified: distributed across all pools (great for availability, long workloads)
  • capacityOptimized: pool with the optimal capacity for the number of instances
  • priceCapacityOptimized (recommended): pools with highest capacity available, then select the pool with the lowest price (best choice for most workloads)
• Spot Fleets allow us to automatically request Spot Instances with the lowest price

Dedicated Hosts

• A physical server with EC2 instance capacity fully dedicated to your use
• Allows you address compliance requirements and use your existing server- bound software licenses (per-socket, per-core, pe—VM software licenses)
• Purchasing Options:
  • On-demand – pay per second for active Dedicated Host
  • Reserved - 1 or 3 years (No Upfront, Partial Upfront, All Upfront)
• The most expensive option
• Useful for software that have complicated licensing model (BYOL – Bring Your Own License)
• Or for companies that have strong regulatory or compliance needs

Dedicated Instances

• Instances run on hardware that’s dedicated to you
• May share hardware with other instances in same account
• No control over instance placement (can move hardware after Stop / Start)

Characteristic	Dedicated Instances	Dedicated Hosts
Enables the use of dedicated physical servers	✅	✅
Per instance billing (subject to a $2 per region fee)	✅
Per host billing		✅
Visibility of sockets, cores, host ID		✅
Affinity between a host and instance		✅
Targeted instance placement		✅
Automatic instance placement	✅	✅
Add capacity using an allocation request		✅

Capacity Reservations

• Reserve On-Demand instances capacity in a specific AZ for any duration
• You always have access to EC2 capacity when you need it
• No time commitment (create/cancel anytime), no billing discounts
• Combine with Regional Reserved Instances and Savings Plans to benefit from billing discounts
• You’re charged at On-Demand rate whether you run instances or not
• Suitable for short-term, uninterrupted workloads that needs to be in a specific AZ

Which option is right for me?

• On demand: coming and staying in resort whenever we like, we pay the full price
• Reserved: like planning ahead and if we plan to stay for a long time, we may get a good discount.
• Savings Plans: pay a certain amount per hour for certain period and stay in any room type (e.g., King, Suite, Sea View, …)
• Spot instances: the hotel allows people to bid for the empty rooms and the highest bidder keeps the rooms. You can get kicked out at any time
• Dedicated Hosts: We book an entire building of the resort
• Capacity Reservations: you book a room for a period with full price even you don’t stay in it

Price Comparison

Example – m4.large – us-east-1

Price	Type Price (per hour)
On-Demand	$0.10
Spot Instance (Spot Price)	$0.038 - $0.039 (up to 61% off)
Reserved Instance (1 year)	$0.062 (No Upfront) - $0.058 (All Upfront)
Reserved Instance (3 years)	$0.043 (No Upfront) - $0.037 (All Upfront)
EC2 Savings Plan (1 year)	$0.062 (No Upfront) - $0.058 (All Upfront)
Reserved Convertible Instance (1 year)	$0.071 (No Upfront) - $0.066 (All Upfront)
Dedicated Host	On-Demand Price
Dedicated Host Reservation	Up to 70% off
Capacity Reservations	On-Demand Price

Placement Groups

• Sometimes you want control over the EC2 Instance placement strategy
• That strategy can be defined using placement groups
• When you create a placement group, you specify one of the following strategies for the group:
  • Cluster—clusters instances into a low-latency group in a single Availability Zone
  • Spread—spreads instances across underlying hardware (max 7 instances per group per AZ)
  • Partition—spreads instances across many different partitions (which rely on different sets of racks) within an AZ. Scales to 100s of EC2 instances per group (Hadoop, Cassandra, Kafka)

Placement Groups Cluster

placement group cluster

• ✅ Pros: Great network (10 Gbps bandwidth between instances with Enhanced Networking enabled - recommended)
• ❌ Cons: If the rack fails, all instances fails at the same time

Use case:

• Big Data job that needs to complete fast
• Application that needs extremely low latency and high network throughput

Placement Groups Spread

placement group spread

• ✅ Pros:
  • Can span across Availability Zones (AZ)
  • Reduced risk is simultaneous failure
  • EC2 Instances are on different physical hardware
• ❌ Cons: Limited to 7 instances per AZ per placement group

Use case:

• Application that needs to maximize high availability
• Critical Applications where each instance must be isolated from failure from each other

Placement Groups Partition

placement group Partition each partition represent a rack in AWS

• Up to 7 partitions per AZ
• Can span across multiple AZs in the same region
• Up to 100s of EC2 instances
• The instances in a partition do not share racks with the instances in the other partitions
• A partition failure can affect many EC2 but won’t affect other partitions
• EC2 instances get access to the partition information as metadata

Use case: HDFS, HBase, Cassandra, Kafka

Hibernate

We know we can stop, terminate instances

• Stop – the data on disk (EBS) is kept intact in the next start
• Terminate – any EBS volumes (root) also set-up to be destroyed is lost

On start, the following happens:

• First start: the OS boots & the EC2 User Data script is run
• Following starts: the OS boots up
• Then your application starts, caches get warmed up, and that can take time!

Overview

• The in-memory (RAM) state is preserved
• The instance boot is much faster! (the OS is not stopped / restarted)
• Under the hood: the RAM state is written to a file in the root EBS volume
• The root EBS volume must be encrypted

Use cases:

• Long-running processing
• Saving the RAM state
• Services that take time to initialize

Good to know

• Supported Instance Families – C3, C4, C5, I3, M3, M4, R3, R4, T2, T3, …
• Instance RAM Size – must be less than 150 GB.
• Instance Size – not supported for bare metal instances.
• AMI – Amazon Linux 2, Linux AMI, Ubuntu, RHEL, CentOS & Windows…
• Root Volume – must be EBS, encrypted, not instance store, and large
• Available for On-Demand, Reserved and Spot Instances
• An instance can NOT be hibernated more than 60 days

EC2 Instance Store

• Provide temporary block level storage for instance.
• The storage is a disk that physically attached to host.
• Good for buffer / cache / scratch data / temporary content
• Ideal workload: buffers, caches, scratch data, temporary content
• Data is lost when:
  • Underlying disk drive fails
  • Instance stop / hibernates / terminates
• Backups and Replication are your responsibility
• Very high IOPS: up to million IOPS (Eg: i3.16xlarge 3.3M read and 1.4M write IOPS). If you need a high-performance hardware disk, use EC2 Instance Store

Ref: Better I/O performance (opens new window)

Amazon Machine Image (AMI)

• AMI are a customization of an EC2 instance
  • You add your own software, configuration, operating system, monitoring…
  • Faster boot / configuration time because all your software is pre-packaged
• AMI are built for a specific region (and can be copied across regions)
• You can launch EC2 instances from:
  • A Public AMI: AWS provided
  • Your own AMI: you make and maintain them yourself
  • An AWS Marketplace AMI: an AMI someone else made (and potentially sells)

AMI Process (from an EC2 instance)

• Start an EC2 instance and customize it
• Stop the instance (for data integrity)
• Build an AMI – this will also create EBS snapshots
• Launch instances from other AMIs

Tạo AMI từ EC2 (opens new window)