Architecting Web Applications for the Cloud - Design Principles and Practical Guidance for Amazon Web Services (AWS)

ARCHITECTING WEB APPLICATIONS FOR
THE CLOUD: DESIGN PRINCIPLES AND
PRACTICAL GUIDANCE FOR AWS
Dr Adnene Guabtni, Senior Research Scientist, NICTA/Data61, CSIRO
Adnene.Guabtni@csiro.au
EC2 S3
ELB RDS AMI AZ
EBS EMR Route53 VPC
Kinesis ElastiCache
DynamoDB CloudWatch
CloudFormation Auto-Scaling Lambda
CloudFront CDN Redshift SES SNS SQS
AWS
EC2 S3
ELB RDS AMI AZ
EBS EMR Route53 VPC
Kinesis ElastiCache
DynamoDB CloudWatch
AWS

LET’S CLEAR ANY MISUNDERSTANDING ABOUT THE CLOUD
29% of the general public think Cloud Technology is an actual cloud (Wakefield Research)
But hey, that’s the general public! What about IT professionals?
Many IT professionals would think that the benefits of the cloud are:
 Portable office.
 Cost Savings.
 Fewer responsibilities, easier manageability.
 Reliability (SLA which guarantees 24/7/365 and 99.99% availability).
Looks like traditional “Hosted services”.
Cloud Computing is Different.
ARCHITECTING WEB APPLICATIONS FOR THE CLOUD: DESIGN PRINCIPLES AND PRACTICAL GUIDANCE FOR AWS © DR ADNENE GUABTNI 2

THE CLOUD COMPUTING DIFFERENCE
 IT Assets Become Programmable Resources
 Servers, databases, storage, and higher-level application components are temporary and disposable,
quickly provisioned when needed.
 They dynamically scale to meet actual demand.
 You only pay for what you use.
Think of how a software allocates memory on demand and “garbage collect” unused objects?
Cloud Computing is similar but applied to virtual resources like servers, databases, storage, …

 No need to know how to program resources
 Rely on a higher level of managed services, such as Auto-scaling, Load balancers, …
 Deliver new solutions faster.
 Designed for scalability and high availability.

 Global, Available, and Unlimited Capacity
 Whether you need to serve 1 user or 1 billion users.
 Whether you need to optimize network
speed for US, Europe, Asia, etc.
 Move machines and data around the
globe programmatically.
 Business Continuity.
 Disaster recovery.
AWS Global Infrastructure (May 2016)

 Global, Available, and Unlimited Capacity
 Security is Built-in
 Native AWS security and encryption features can help achieve higher levels of data protection and
compliance.
 Security policies built-into programmable resources.
 Continuous monitoring of configuration changes to your IT resources.
 Auditing is no longer periodic or manual, it becomes part of your continuous delivery pipeline.

AWS ARCHITECTURE DIAGRAMS
AWS architecture diagrams are a great way to communicate about your design,
deployment and topology. In the following slides, the official collection of AWS
Simple Icons v2.4 is used. These include:
 Compute & Networking
 Storage & Content Delivery
 Database
 Enterprise Applications
 Administration & Security
 Deployment & Management
 Application Services
 Analytics
 Mobile Services
 Non-Service Specific
 On-Demand Workforce
 SDKs
 Groups

 Database
 Analytics
 Mobile Services
 SDKs
 Groups

TOOLS

TOOLS
creately.com

TOOLS
cloudcraft.co

DESIGN PRINCIPLES FOR AWS
SCALABILITY
Scaling Vertically
 1 single EC2 instance
 Type: M3
 Size: Medium

SCALABILITY
Scaling Vertically
 Type: M3
 Size: Medium
 COST: $48.24/month
Cheap but not large enough.

SCALABILITY
Scaling Vertically
 Type: M3
 Size: Medium
Selecting higher specification.

SCALABILITY
Scaling Vertically
 Type: M3
 Size: Xlarge
 COST: $191.52/month
Still cheap but there is a maximum capacity.
(2Xlarge for M3-General instances)

SCALABILITY
Scaling Horizontally
 Multiple EC2 instances
 Type: M3
 Size: Medium

SCALABILITY
 Type: M3
 Size: Medium
+ Elastic Load Balancer
Unlimited capacity,
simply add more instances.

SCALABILITY
 Type: M3
 Size: Medium
What about Databases?

SCALABILITY
 Type: M3
 Size: Medium
+ Relational Database Service
(RDS)

SCALABILITY
A full example of Scalable Web Application Architecture
 Total minimum cost : $980.64/month
That’s a good price for
a fully scalable deployment
that can serve 1 user
or 1 billion users*
* cost would sky rocket when scaled to serve 1 billion users

SCALABILITY AT ALL LEVELS
Scalable CDN
Content Delivery Network using AWS CloudFront.
Scalable DNS
Domain Name System web service using AWS Route 53.
Scalable Load Balancer using ELB
Scalable Front end component
Elastic Compute Cloud (EC2) with Auto Scaling
Scalable API component
Elastic Compute Cloud (EC2) with Auto Scaling
Scalable DB
RDS with multi-AZ Replication
Scalable storage for static files
Simple Storage Service (S3)


LOOSE COUPLING
Statefull components as services
 Example: Front end.
 Service Discovery via Elastic Load Balancers
as stable endpoints.
 Use of Sticky Sessions.
 Load balancers implement the Push model.
Stateless components as services
 Example: Backend API.
 Load balancers (Push Model) can be used
but not best
 Ideally use SQS (Simple Queue Service)
for Asynchronous Integration (Pull Model).


CHOSE THE RIGHT DATABASE(S)
Relational Databse
 RDS is a MySQL compatible databse that offers:
 Scalability (both vertically and horizontally).
 High Availability (RDS Multi-AZ deployment).
Automatic failover to the standby without the need
for manual administrative intervention.
NoSQL Databases
 DynamoDB is a NoSQL database.
 Suitable if your application primarily indexes
and queries data with no need for joins
or complex transactions.
 Scale both the reads and the writes in a horizontal fashion.
 High Availability: synchronously replicates data across three
facilities in an AWS region to provide fault tolerance.


CHOSE THE RIGHT DATABASE(S)
Data Warehouse
 Specialized type of relational database, optimized for
analysis and reporting of large amounts of data.
 Redshift is a managed data warehouse service.
 SQL-based.
 Scalability using massively parallel processing (MPP),
columnar data storage, and targeted data
compression encoding schemes.
 High Availability with multi-node clusters
in which data written to a node is
automatically replicated to other nodes
within the cluster.
+ backed up to Amazon S3.

Amazon Redshift

DISPOSABLE RESOURCES INSTEAD OF FIXED SERVERS
Auto Scaling allows EC2 instances to be discarded
when not used and new instances provisioned
in seconds to meet the demand.
RDS database scales into several nodes
and any node can be discarded
without impact on DB availability.
Everything else is a service.


REMOVING SINGLE POINTS OF FAILURE
Automated Multi-Data Center Resilience
 Each AWS region contains multiple distinct locations
called Availability Zones (AZ).
 Each AZ is engineered to be isolated
from failures in other AZ.
 If AZ “A” fails, failover is automated
and all requests are routed to the
working AZ “B”.
Note: Multiple AZs and multiple regions can be used.


SERVICES, NOT SERVERS (AS MUCH AS POSSIBLE)
Use Managed Services
 Simple Queue Service (SQS).
 CloudFront for content delivery.
 Elastic Load Balancers.
 RDS.
 DynamoDB.
 CloudSearch.
 Simple Email Service (SES).
 S3.
 …


SERVICES, NOT SERVERS (100%)
Serverless Architectures
Example: Voting mobile app.
 Voting app on mobile
uses Amazon API Gateway to vote.
 Vote request is routed
to an AWS Lambda function (compute service).
 AWS Lambda function extracts info
and record it into DynamoDB (NoSQL service).
 DynamoDB triggers anAWS Lambda function
to generate static HTML and store it
into S3 (file storage service).
 Route 53 (DNS service) serves the static HTML content
from S3 to the user.


CACHING
Application Data Caching using AWS ElastiCache
 In-memory caching engines:
 Memcached (objects).
 Redis (key-value store).
 Automatically detects and replaces failed nodes.
 Faster than disk.
 Scale vertically or horizontally.
Edge Caching using CloudFront
 Content Delivery Network consisting of multiple
edge locations around the world.
 Edge caching allows content to be served by infrastructure
that is closer to viewers.


OPTIMIZE FOR COST
Select the right type/size for your instances.
 General Purpose vs Compute Optimized vs Memory Optimized vs Storage Optimized.
 Burstable Performance Instances or fixed performance instances.
 Large vs Xlarge vs 2Xlarge vs 4Xlarge vs 10Xlarge.
 Many small instances vs fewer large instances.
Rely on Auto-Scaling to always fit the demand and pay for what you need.
Take Advantage of the Variety of Purchasing Options.
 Reserved Capacity.
 On-Demand vs Spot Instances (Bidding strategy).
 Mix On-Demand and Spot Instances.


SECURITY

Use VPC (Virtual Private Cloud).
A logically isolated section of Amazon Web Services (AWS) Cloud.
Use a Bastion Host.
A special purpose server instance that is
designed to be the primary access point
and acts as a proxy to your other
EC2 instances.
Use IAM (Identity and Access Management)
to define a granular set of policies and assign them
to users, groups, and AWS resources.

SECURITY

For Web Applications.
Use WAF (Web Application Firewall).
Protects against SQL injection and other vulnerabilities.
Setting up SSL on an Elastic Load Balancer.
Allows to offload your instances from managing
SSL encryption/decryption.
Your SSL certificates are safe, within ELB,
not within your instances.
Cypher suite configuration is always up to date,
upgraded by Amazon when necessary
(in case of new vulnerability).

SUMMARY OF DESIGN PRINCIPLES FOR AWS
 Scalability at all levels.
 Loose Coupling.
 Chose the right database(s).
 Disposable Resources Instead of Fixed Servers.
 Removing Single Points of Failure.
 Services, Not Servers (as much as possible).
 Caching.
 Optimize for Cost.
 Security.

LEARN MORE AND KEEP UP TO DATE
White Paper: Architecting for the Cloud - AWS Best Practices
https://aws.amazon.com/whitepapers/architecting-for-the-aws-cloud-best-practices/
Amazon regularly introduces new services, so keep up to date
https://aws.amazon.com/new/
AWS Free Tier (12 months free to get started)
https://aws.amazon.com/free/

THANK YOU FOR YOUR ATTENTION Dr Adnene Guabtni
Adnene.Guabtni@csiro.au
EC2 S3
ELB RDS AMI AZ
EBS EMR Route53 VPC
Kinesis ElastiCache
DynamoDB CloudWatch
AWS
EC2 S3
ELB RDS AMI AZ
EBS EMR Route53 VPC
Kinesis ElastiCache
DynamoDB CloudWatch
AWS
slideshare.net/guabtni
Access these slides on

Architecting Web Applications for the Cloud - Design Principles and Practical Guidance for Amazon Web Services (AWS)

Recommended

More Related Content

Similar to Architecting Web Applications for the Cloud - Design Principles and Practical Guidance for Amazon Web Services (AWS) (20)

Recently uploaded (20)

Architecting Web Applications for the Cloud - Design Principles and Practical Guidance for Amazon Web Services (AWS)