This document discusses scalability, load balancing, and auto scaling on AWS. It defines vertical and horizontal scalability. Load balancers distribute traffic across instances and ensure high availability. Auto scaling groups dynamically add or remove instances based on metrics to maintain performance during varying loads.
This document discusses scalability, load balancing, and auto scaling on AWS. It defines vertical and horizontal scalability. Load balancers distribute traffic across instances and ensure high availability. Auto scaling groups dynamically add or remove instances based on metrics to maintain performance during varying loads.
This document discusses scalability, load balancing, and auto scaling on AWS. It defines vertical and horizontal scalability. Load balancers distribute traffic across instances and ensure high availability. Auto scaling groups dynamically add or remove instances based on metrics to maintain performance during varying loads.
This document discusses scalability, load balancing, and auto scaling on AWS. It defines vertical and horizontal scalability. Load balancers distribute traffic across instances and ensure high availability. Auto scaling groups dynamically add or remove instances based on metrics to maintain performance during varying loads.
WWW.CYBERFRAT.COM Scalability & High Availability 2
• Scalability means that an application / system can handle greater loads by
adapting.
• There are 2 types of Scalability
• Vertical Scalability. • Horizontal Scalability
WWW.CYBERFRAT.COM Vertical Scalability 3
• This means increasing the size of the instance.
• E.g scaling from t2.micro to t2.large. • Commonly used to non-distributed systems such as db. • RDS, ElastiCache also can use vertical scaling.
WWW.CYBERFRAT.COM Horizontal Scalability 4
• It means Increasing the number of instance/systems for app.
• It means the system would be distributed. • Very Common for web apps/ containerized applications. • Easy to scale horizontally .
WWW.CYBERFRAT.COM High Availability 5
• It means running your application /system
in at least 2 availability zone (AZ)
• If the application in one AZ is down we
still have it on one AZ.
• This one goes hand in hand with
horizontal scaling
WWW.CYBERFRAT.COM Vertical Scalability 6
• This means increasing the size of the instance.
• E.g scaling from t2.micro to t2.large. • Commonly used to non-distributed systems such as db. • RDS, ElastiCache also can use vertical scaling.
WWW.CYBERFRAT.COM What is load balancing? 7
• Load balancers in simple words are devices/servers which will
forward the traffic to multiple servers attached to it.
WWW.CYBERFRAT.COM Why to use load balancer? 8
• To distribute the load across multiple servers/instance attached to it .
• You can expose your application to a single point of access (DNS). • It handles the failures of the servers/instance attached to it ( won't send traffic on those servers). • It does a regular health checks to your servers/instance ( we have to provide the inputs). • Enforces session stickiness with cookies. • High availability across multiple availability zones.
WWW.CYBERFRAT.COM Why use an EC2 Load balancer ? 9
• An ELB (Ec2 Load Balancer) is an AWS managed load balancer. We get is as a
service. • AWS assures us the uptime of the Load balancer, we will not have to worry about that. • AWS takes care of it's availability, upgrades & maintenance, we do not have to worry about that. • It is integrated to your AWS account as a service.
WWW.CYBERFRAT.COM Health Checks 10
• Health checks are very import part of a Load balancer.
• With health checks the LB knows if the server attached to the load balancer are available or are in healthy state. • Health check is done on a port and a specific route or path. • If response if not 200(OK), then the instance is unhealthy.
Health checks
WWW.CYBERFRAT.COM Types of Load Balancers on AWS 11
• Currently AWS has 3 types of Load balancer those are managed by them. • Classic Load balancer (v1 Old generation) - 2009 • HTTPS, HTTP, TCP • Application Load balancer (v2 new generation ) - 2016 • HTTPS, HTTP, WebSocket • Network Load balancer (v2 new generation ) - 2017. • TCP, TLS(secured TCP) & UDP
When creating a load balancer you can select if the load balancer is internal facing or public facing
WWW.CYBERFRAT.COM Classic Load balancer 12
• Works on layer 4(TCP) & layer 7( HTTP & HTTPS)
• Health check are TCP & HTTP based. • They have fixed hostname • E.g - mylb.region.elb.amazonaws.com
listener internal
Client CLB EC2
WWW.CYBERFRAT.COM Application Load balancer 13
• It is a layer 7 load balancer (HTTP).
• Load balancing on multiple HTTP based application on multiple servers/instances ( Target group). • It can also load balance to multiple application on same instance/server. ( e.g Containers). • It also supports redirection (HTTP to HTTPS),
WWW.CYBERFRAT.COM Application Load balancer 14
• It routes traffic to multiple target groups:
• path based routing from URL ( amit.com/users) • hostname based routing (one.amit.com , app.amit.com) • Query string, headers based routing ( amit.com/user?id=123&order=false) • ALB is a good fit for microservers based application (docker or ECS). • It can also load balance to multiple application on same instance/server. ( e.g Containers). • port mapping feature to redirect to dynamic ports in ECS
WWW.CYBERFRAT.COM Application Load balancer 15
Target group for
USER application
amit.com/user
www
Target group for
SEARCH application www amit.com/search
WWW.CYBERFRAT.COM Network Load balancer 16
• Layer 7 load balancer.
• It forwards UDP & TCP traffic to the instances • Less latency ~100ms (400ms for ALB) • NLB is having 1 static IP per AZ & we can assign ElasticIP. • NLB are used for extreme performance, TCP or UDP traffic
WWW.CYBERFRAT.COM Load Balancer Stickiness 17
• If Stickiness is enabled the same client's request will be always redirected to same instance behind load balancer. • This option is available for only classic & Application Load balancer. • We can choose the cookie expiry time. • Useful when we have make sure the user will not lose his session.
WWW.CYBERFRAT.COM SSL/TLS - Basics 18
• With SSL certificate we can have the traffic between the clients & load balancer to be encrypted in transit (in-flight). • SSL - Secure Socket Layer • TLS - Transport Layer Security, Newer version than SSL • These SSL Certificates are issued by CA ( Certificate Authorities) • e.g GoDaddy, Digicert, Symantec etc.. • You have to renew the SSL certificates as they have expiry dates. • LB uses X.509 certificates ( SSL/TLS server certificates) • You can manage all the certificates in Amazon's certificate manager called as AWS Certificate Manager. • We can also create our own certificate (self-signed) ls upload to ACM. WWW.CYBERFRAT.COM What’s an Auto Scaling Group? 19
• In real time environment the traffic to your website or Application can change at any point of time.
• Why use the Auto Scaling group (ASG) ?
• Scale out (add EC2) to match the high load/request. • Scale in (remove Ec2) to match the low load/request • Making sure we have a minimum & maximum number of machines running. • It can automatically register new instances to a load balancer.
WWW.CYBERFRAT.COM Auto Scaling Group in AWS ? 20
WWW.CYBERFRAT.COM Auto Scaling with AWS Load Balancer ? 21
WWW.CYBERFRAT.COM ASGs have the following attributes: 22
• A Launch configuration. • AMI + instance type • Ec2 User data • EBs volumes • Security group • SSH key pair ● Min Size / Max size / Initial capacity ● Network + subnet Information ● Load balancer information ● Scaling Policies
WWW.CYBERFRAT.COM Auto scaling Alarms. 23
• Possible to scale based on Cloudwatch alarms.
• An Alarm monitors metrics likes (cpu, network, etc). • Metrics are calculated for over all ASG instances. • Using these alarms we can: • create scale out policy (adding instances) • create scale in policy ( removing instances)
WWW.CYBERFRAT.COM Auto scaling Custom metric. 24
• We can auto scale based on custom metric
• We have to send custom metric from application on EC2 to AWS Cloudwatch • Using cloudwatch Alarms to react to low/high values • Using the created alarm as scaling policy for ASG.
