AWS Certified Solutions Architect Associate - Part 7 - Autoscaling and virtual network services

March 09, 2020

Auto Scaling Solutions

Overview

Ability to automatically scale out or in according to our needs at a point in time.

Monitors apps
Adjusts capacity

How Auto Scaling Works

Scalable Resources

EC2 Auto Scaling Groups
Aurora DB clusters
DynamoDB global secondary indexes
DynamoDB tables
Elastic Container Service (ECS)
Spot fleet requests

Auto Scaling Costs

Free to use
Results of use may cost:
- More instances
- CloudWatch
- ELB load balancers

You could limit total amount spendable per hour due to auto scaling.

Scaling Groups

A collection of instances with similar characteristics
- Can be scaled based on criteria like network comms, CPU utilization
- Unhealthy instances can be automatically replaced. Any state other than running
is considered unhealthy

Group Considerations

Time to launch and configure server is required, to be able to scale out
Relevant metrics to your application
- CPU utilization
- Network throughput
- Free memory
What AZs should the auto scaling group span. Recommended to have multiple

instances over different AZs.

Scale to increase or decrease capacity
Specify the minimum number of instances always running

Termination Policies

Select AZ with most instances
If more than one of above, select AZ with instances with oldest launch config
if more than one instance with oldest launch config, select instance closest to

the next billing hour

Auto scaling of instances may rely on: Number, Age, Cost

Auto Scaling Termination Policy

Custom Termination Policies

OldestInstance - likely to need overhaul
NewestInstance
OldestLaunchConfigurations
ClosestToNextInstanceHour - likely to reduce cost
Default

Auto Scaling Configuration

Configuration for how instances added to the auto scaling group will be initiated
When Creating and instance, you could choose to create more than one and launch

them in an auto scaling group

A launch configuration is needed to create an auto scaling group. It can be

created before or during the creation of the auto scaling group.

The launch configuration contains the instance type, key pair, security groups etc

Steps to create an auto scaling group

Management Console -> EC2
Auto scaling on the left -> Auto scaling groups -> Create auto scaling group
Create or select a launch template
AMI. You could create a new AMI or from an existing instance.

Create Launch Configuration

Name
Request spot instances
IAM role - Role you want associated with this instance
CloudWatch
Advanced details
- Kernel ID
- RAM disk ID
- Bootstrap - linux bash scripts for startup
Assign Elastic IPs (EIP) for any instances you want to be publicly accessible

over internet

If using load balancer then no need to worry about EIPs.
Add security group to enable e.g HTTP, SSH, SMTP etc
Review
New or existing key pair
Click create launch configuration

Create Auto Scaling Group

Minimum instances in the auto scaling group
Network
Subnet - Choose all/multiple subnets in your VPC to allow autoscaling across

multiple or all AZs -> High availabiity.

Advanced details, Enable load balancing etc.
Select use scaling policies to adjust capacities
- Metric - CPU usage
- Value - 80 -> 80%
Review
Create

Launch Methods

Launch template, prerequisites
- The template must include all the required parameters need to launch an EC2
instance otherwise error: You need to use a fully formed launch template
- IAM user/role must have permission for EC2:run instances permission
Launch config
Existing EC2 instance
EC2 launch wizard. When you create an EC2 instance, there is an option

to create more than one instance with a link to put that instance in an auto scaling group

Load Balancer Concepts

Load balancer sits between the requesters and servicers (which respond to requests) Load balancer takes the incoming request and decide which servicer will handle.

Load Balancing Categories

Sender initiated. Sender locates the best target e.g DNS load balancing. DNS sends names of

multiple names and the client selects the best target.

Receiver initiated. Receiver selects best target.

Classes of Load Balancing

Static Load Balancing used by multi-tier application
Dynamic load balancing - true load balancing
- Actions dynamically assigned
- Scalability is provided by this load balancing model

Elastic Load Balancer (ELB) uses dynamic load balancing.

Load Balancing Algorithms

Round robin - serial allocation of requests to servicing nodes
Randomized
Centrally managed - Decisions made based on available info on the environment
Theshold based - Send request to on servicing node up to a certain amount

before switching to the next servicing node

Remember all nodes need to have same data

replication
storage arean network

Elastic Load Balancing (ELB)

ELB Benefits

Highly available
Secure (Security handled by AWS)
Flexible
Monitoring and auditing included
Elastic
Hybrid - Multiple types of load balancing could be associated

ELB Types

Classic Load Balancer. Not recommeded for any newly deployed instances
Application Load Balancer - Web app, HTTP level (OSI model layer 7)
Network Layer - TCP level (OSI model layer 4)

Supported Services

EC2
ECS
Auto Scaling
CloudWatch
Route53

Read Amazon AWS Documentation

Elastic Load Balancing Features -> Product Comparison

Virtual Network Services

Domain Name System (DNS)

Domain Name System (DNS) - Name, location service for IP networks. DNS helps resolve human friendly names to actual IP address.

RFC 1034 and 1035
Provides name to IP address mapping

Fully Qualified Domain Name (FQDN)

Format: .<management/organization domain>.
Example: aws.amazon.com
Domain = . = amazon.com
Example with sub domain: services.aws.amazon.com

Requestor -> root DNS server -> top level domain DNS Server -> Specific DNS Server Web browser -> root DNS serve -> com server -> AWS DNS Server

Requestor could be device, server, DNS server etc

DNS Hosting

Provide name resolution with:
- Caching
- Recursion - recursively looksup till arriving at the required IP address
Stores DNS database which supports aliasing
Provides DNS zone transfers
- Offloads name resolution processing

DNS Resolution

Requests IP address of a host name (forward lookup)
Request host name of an IP address (reverse lookup)
Requestor is configured with a DNS server address for example Google public

DNS servers 8.8.8.8 and 8.8.4.4

DNS Records

A and AAAA
- Used to resolve hostname to IP address (A for IPv4, AAAA for IPv6)
NS records (Name server records)
- Domain to hostname resolution
MX record - Mail record
CNAME record
- Alias for real name (canonical)
SRV record - service records

Configuring DNS

Customize IAM login link

Management Console -> IAM
At the top you see the login link
Click customize to give it an alias
Alias must be globally unique

Each VPC as a DNS configuration set. AWS generates a host name and public IP address for EC2 instances etc. The host name cannot be changed but another host name could be associated with the public ip address using Route53 or other DNS Services.

Configuring Route 53

Managed (you don’t have to create an instance) Service within AWS
Management Console -> Networking & Content Delivery -> Route53
DNS Management
Hosted Zone (Info on traffic routing) -> Create Hosted Zone
- Hosted Zone is like an internal private DNS
- We create a hosted zone for routing traffic within our VPC.
Specify a domain name.
- Domain names are resoloved within the VPC. If the domain names you specify already
exists publicly (internet) then the conflict may prevent users of your VPC from browsing to the already existing public domain.
Specify your VPC
Click create
After creation it shows you (SOA) the dns server under which you are and also

.com, .org and .co.uk servers.

You can mirror another internet network into yours by importing their zone and

using the same subnets and ip addresses.

Routing Example using 2 IP addresses and clustered load balancing techniques

Management Console -> Networking & Content Delivery -> Route53
Create a Record
Select Type: A - Ipv4
Values -> Enter 2 ip addresses
Routing Policies
- Simple traditional DNS
- Weighted - Go to one IP 80% and another IP 20% of the time. Useful when
deploying a new version of an app etc
- Latency - Send users to the destination with the lowest latency
- Failover - Use one ip address until it fails then use the second
- Geo location - Get the location of the user and forward to closest server
- Multi value answer -

Health Check

Protocol
IP address
Host name
port
pat

Configuring ACLs and NACLs

ACL

Management Console -> s3
Click on a bucket
Click on permissions

NCL

Management Console -> VPC
Click on Network ACL

Create Network ACL

Name
VPC
Create

After creating the ACL

Set inbound rules.
- Rules are processed for the top of the list and once a match applys stops.
Set outbound rules
- Rules are processed for the top of the list and once a match applys stops.
- Default denys all access by default.
- Add a rule

Working with ACL/rules at the subnet level

Mehod A