Deploying microservices on AWS

DEPLOYING
MICROSERVICES
ON AWS
MICHAEL HABERMAN
FOUNDER @ APPLICALLS

TODAY
What is microservices?
How to host microservices?
Deployment aspect of
microservices

GENERAL CONCEPT
Instance
User logic
Payment logic
Analytics logic
Reporting logic
Monolith
User logic
Payment logic
Analytics logic
Reporting logic

WHY IT IS GOOD FOR US?
• Large project are hard to maintain
• Easy to test
• Safe to deploy
• Can use different programming languages
• Can use different databases

WHY IT IS GOOD FOR US?
Who is us…?

WHY IT IS BAD FOR THEM?
• Multiple instance to handle
• Complicated deployment
• A lot of automation
• So many logs…
• Complicated monitoring and alerts
• Configuration….

WHY IT IS BAD FOR THEM?
Who is them…?

TODAY
• What is microservices?
• How to host microservices?
• Deployment aspect of microservices

HOSTING OPTIONS IN AWS
• EC2 – Elastic Cloud Compute
• Elastic beanstalk
• ECS – Elastic Container Service (Docker…)
• Lambda Function - Serverless

EC2
• Simple VM in the cloud
• Image as base installation
• Manual maintenance (updates, backup)
• Elastic beanstalk is fully managed
• Auto Scaling defined by you
• Pay per hour
• Ideally instance per application / service

EC2 – CLOUD DIAGRAM
Load
balancer
Service 1
Service 2
EC2
EC2
EC2
EC2

EC2 - PROS AND CONS
• Pros
• Easy to migrate
• Better for different OS (Linux and Windows)
• Cons
• Expensive (Compute & storage)
• Hard to maintain
• Auto scaling may take time

Docker Host (EC2)
DOCKER
Docker Host (EC2)
Container 1
Container 2
Container 3
Container 4
Container 3
OS
Web Server
App code

ECS
• Managed Docker by AWS
• Container ”image” as Code (Docker)
• Cluster image managed by AWS
• Two level auto scale, cluster & container
• High resources utilization
• Container == service (or in javascript container === service)

ECS – CLOUD DIAGRAM
Load
balancer
EC2 #1
EC2 #2
Service 1
Service 2
Service 1
Service 2

ECS - PROS AND CONS
• Pros
• Easy to move from AWS
• Easy to run locally
• Good utilization of resources
• Easy and fast auto scale
• Easy to maintain the containers
• Cons
• Another technology knowledge (Docker)
• Windows Docker container is problematic

LAMBDA FUNCTION
• Serverless – no underlying infrastructure
• No maintenance
• Pay only of runtime
• Focus on the code

LAMBDA – CLOUD DIAGRAM
API Gateway
Lambda 1
Lambda 2

LAMBDA - PROS AND CONS
• Pros
• Cheap at low – medium scale
• Much less maintenance
• Fully managed by AWS (scaling)
• Cons
• Hard to develop locally
• Specific programing languages
• No file system
• Limited installations
• Not mature enough

COMPARISON
EC2 ECS Lambda
Maintenance High Medium Low
Cost Effective Low High Depends on scale
Flexibility Low High High
Auto scaling A lot of work Some work Managed by AWS
Coupled with AWS None Low High
Local development Easy Easy Complex

DEPLOYMENT
• Source control
• Build
• Test
• Deploy
• Test & rollback
• Monitor
• Alerts

SOURCE CONTROL
• What do we need:
• Allow multiple developers to work in parallel
• Easy way to merge
• Easy way to get specific version code
• Support multiple environments

SOURCE CONTROL
GIT FLOW ENV BASED
Master Dev Feature

LAMBDA - PROS AND CONS
What would you choose?

SOURCE CONTROL
• Why git flow?
• Docker image should be build once
• Same image moves between envs freely
• Use tags to indicate image usage
• Don’t rebuild images between envs
• It is a new app (npm install….)

BUILD
• Independent build server!
• Jenkins, CodeBuild
• Don’t set configuration at this point
• Build should include tests (more on that later)
• Push to repository with matching tag (latest stable build)

TEST
• Static
• No need to run the service, but specific parts
• Dynamic
• Need to run the service, but doesn’t communicate
• Depended dynamic
• Need to run the service and it’s dependencies

DEPLOY
• Deploy single service vs Entire cluster
• Retag
• Live testing
• rollback

Docker Host (EC2)
SINGLE CONTAINER DEPLOYMENT
Docker Host (EC2)
Container 1
Container 2
Container 3
Container 1
Container 2
Container 3
Container 3 new Container 3 new
Traffic
routed to
new
Killed
after no
traffic
Tested

Docker Host (EC2)
CLUSTER DEPLOYMENT
Docker Host (EC2)
Container 1
Container 2
Container 3
Container 1
Container 2
Container 3 new
Traffic
routed to
newKilled
after no
traffic
Tested

DEPLOY SINGLE SERVICE VS ENTIRE CLUSTER
• Entire Cluster
• Easier to deploy few services together
• Single Container
• Simpler

LIVE TESTING
• Once deployed
• Run E2E / API testing
• Rollback if needed

MONITORING
• Application monitoring
• Logs
• Configuration
• Infrastructure monitoring
• Logs
• Metrics
• Automatic actions
• Cost

ALERTS
• Based on metrics and events
• Should support: Email, SMS, Phone calls, Slack, and APIs
• Should support alerts based on logs

THANK
YOU!
MICHAEL@APPLICALLS.IO

Deploying microservices on AWS

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Deploying microservices on AWS

Similar to Deploying microservices on AWS (20)

More from Michael Haberman

More from Michael Haberman (14)

Recently uploaded

Recently uploaded (20)

Deploying microservices on AWS