Tuesday, January 3, 2012

Architecting an Highly Available and Scalable WordPress Site in AWS

Why this presentation ?
  • WordPress is used by over 14.7% of Alexa Internet's "top 1 million" websites
  • As of August 2011 WordPress powers 22% of all new websites
  • As of December 2011, version 3.0 had been downloaded over 65 million times
  • Provide some architectural insights to build Highly Scalable and Available WordPress sites in AWS
  • Help startups and companies who are new to AWS- WordPress with some blueprints


FAQ for this architecture can be found here

WordPress Architecture Requirements in AWS
  • Ability to Serve millions of Pages per day
  • Ability to scale out/down dynamically depending upon traffic
  • Ability to perform well even with scaling demands 
  • Should be Highly available without Single Points of Failure
  • Automated Operations – Backup and Monitoring
  • Cost effective 
Why AWS is better for WordPress Scaling?
  • AWS provides flexibility to Scale up and Scale out
  • AWS provides Building block services which is inherently robust and fault tolerant
  • AWS provides full range of options from CDN , OS flavors , Storage Pool , Database , Load Balancers , Monitoring etc
  • Pay as you go – Save cost by leveraging Auto Scaled WordPress EC2 instances
  • We can start small and grow big in AWS

AWS building blocks used
  • Amazon Route53 for managing and answering DNS queries
  • Amazon Elastic Load Balancer for load balancing HTTP/S requests to WordPress Instances
  • RDS MySQL Database Server for storing the WordPress content
  • Simple Storage Service (S3) for storing the WordPress application assets , snapshots and log files
  • Amazon CloudFront CDN for delivering the application static assets from nearest edge locations
  • Amazon CloudWatch for monitoring and sending SNS alerts
Highly Scalable Architecture for WordPress application



Load Balancing Tier
  • Amazon Elastic Load Balancing will Load balance the http requests to WordPress EC2 instances
  • ELB will direct the requests to WordPress EC2 across Multiple Availability Zones ( for HA)
  • Health check , LB algorithms and SSL termination will be done by the Load Balancer
  • Amazon Elastic Load Balancing can keep expanding its capacity automatically depending upon the traffic 
  • Amazon ELB works with Route 53 and Auto Scaling seamlessly
  • Amazon ELB is priced @“ Pay for use” model
  • Amazon ELB is a managed service , so No Maintenance headaches , SW/HW upgrades ,Capacity planning  and manual intervention for expansion
How scalable is this architecture ?
  • Scalability and Elasticity is built in most of the layers in this architecture
  • WordPress EC2 instances can be scaled out and down depending upon the traffic 
  • We can expand the number of WordPress EC2 instances from 1 to 100+ automatically during load peaks
  • We can reduce the number of WordPress EC2 instances from 100+ to 1 automatically during valleys
  • Can serve millions of pages with ease 
  • Pay only for the EC2 instance hours used
  • Read Scalability is built in the Database layer with RDS Read Replicas
  • New RDS Read Replicas can be added / removed with ease
  • Tips:  During Campaigns alone we can run more RDS Read Replicas , Not more than max 5 RDS Read replicas are recommended , Keep Read Replica’s and RDS Master same EC2 size for better performance.
  • Clustered + Distributed File System of the Storage pool layer can be manually scaled in case needed
  • Minimum 2 EC2 Large instances should be used for Storage Pool Layer ( for HA and better IO)
  • Design the Storage Pool Layer with HA (very critical)
  • AWS building blocks like S3 , CloudFront , CloudWatch , SNS used in this architecture  are inherently designed for scalability

How High Availability and Fault tolerance is built in this Architecture ?
  • HA @ WordPress Layer
  • Multiple WordPress EC2 instances avoid single point of failure
  • WordPress EC2 instances are launched across multiple –AZ’s inside a region for High Availability
  • HA @ DB Layer
  • RDS MySQL Master and Standby are launched in 2 different availability zones for High Availability
  • RDS Read Replicas are created in Multiple –AZ’s
  • HA @ Storage Pool Layer
  • GlusterFS is used for the Storage Pool Layer
  • Two EC2 instances used for Storage Pool Layer
  • Storage Pool is setup in replicated mode for High Availability 
  • AWS building blocks like S3 , Cloud Front , CloudWatch , ELB , SNS ,EBS used in this architecture are inherently designed for Fault tolerance and HA
How the Database Tier is Architected ?
  • RDS Master and Standby provide High availability
  • RDS Read Replicas provide Read performance
  • HyperDB plugin is configured to use multiple endpoints like RDS master and RDS read replicas in this architecture
  • DB security groups will allow DB access only to WordPress EC2 instances
  • Periodic Dumps , Snapshots and Point in time recovery is possible in this architecture
What performance aspects are taken care in this Architecture ?
  • APC (or) Xcache plugin can be used for PHP opscode caching
  • W3TotalCache+CloudFront (or) BatCache+Memcached can be used for Page caching
  • Multiple RDS Read Replicas for Read performance
  • RDS Master and Read Replicas are separated to get independent write and read performance
  • Amazon ELB + Auto Scaling improves the overall Site performance when the load increases
Content Delivery Network
  • Amazon CloudFront will be the Content Delivery Network (CDN) 
  • W3TotalCache plugin will be configured to use Amazon CloudFront for enhanced performance and reduced latency
  • Static assets , templates , themes , images etc will delivered from the nearest edge locations of the CDN
Deployment
  • Distributed File Storage Pool is configured between WordPress Management and Content instances
  • WordPress Management and Content instances share the common storage pool for files and plugins
  • Deployment of files and plugins will happen through the WordPress management node
  • Files and plugins will be immediately available for use in the auto scaled WordPress EC2 instances
Monitoring and Alerts
  • Amazon CloudWatch will monitor the CPU and Network utilization of the entire setup
  • Amazon CloudWatch alarms configured with Amazon SNS provides Email/SMS alerts to System Administrators
Backups
  • Custom ops scripts will backup the necessary files from Storage pool to S3 periodically
  • S3 can be configured to remove old backups automatically
  • RDS MySQL is configured to take periodic data dumps and DB snapshots
  • RDS layer can be recovered point in time from the backups

Security
  • Suitably hardened OS for WordPress Mgmt and WordPress instances
  • Firewall (or) AWS security groups configured between all the layers in the architecture
  • SFTP/FTP access only to the WordPress management node
  • AWS IAM policies to manage user account access
  • Install WordPress Security plugins 

Advantages
  • Highly Available and Scalable architecture
  • Can elastically scale out to serve millions of hits in a day
  • Can grow with load demands in future
  • Usage of inherently fault tolerant AWS building blocks adds Stability
  • High performance using CDN and suitable cache plugins
  • Monitoring , Backup and Recovery is built in
  • Pay for use 

Disadvantages
  • Complex to setup and maintain
  • Will not be ideal for smaller sites that do not need scale
  • Will not be cost efficient for sites that have less traffic
Key Points to Remember 
  • Log files generated in WordPress EC2 instances have to rotated 
  • Use Amazon EBS for Storage Pool and WordPress EC2 instances 
  • Start with EC2 m1.Large Instances for the storage pool and WordPress instances
  • More Memory – Better Performance for DB
  • Start with RDS Large for Master 
  • Keep RDS Master and RDS Read Replica’s in same size to improve read performance
  • RDS MySQL supports only Innodb engine
  • Separate the WordPress Management and WordPress Content instances for scalability
  • Do not Scale out/down rapidly in a hour, it will cost more in AWS
  • Combine AWS On-Demand and Reserved Instance pricing to get more savings per month
  • Amazon Availability Zones(AZ’s) are distinct physical locations with Independent power , cooling ,network and security having Low latency network connectivity between them inside the same region
  • Leverage them in WordPress , Storage Pool and DB layers for HA as mentioned in the architecture 

2 comments:

Big Data Events said...

Nice article, responsive themes are useful for attracting users.

Anonymous said...

Hi,

I came across your article while trying to setup our autoscaled system. We are trying to use GlusterFS for File pool.

But, we are facing one issue when we bring up a new instance and try to add it to GlusterFS cluster.


When we try to add the new instance to cluster by issueing command gluster peer probe server1 we get response as server1 is already part of another cluster.

Do you any suggestion on how to add the new instance to the exiting cluster?

Thanks.

Need Consulting help ?

Name

Email *

Message *

DISCLAIMER
All posts, comments, views expressed in this blog are my own and does not represent the positions or views of my past, present or future employers. The intention of this blog is to share my experience and views. Content is subject to change without any notice. While I would do my best to quote the original author or copyright owners wherever I reference them, if you find any of the content / images violating copyright, please let me know and I will act upon it immediately. Lastly, I encourage you to share the content of this blog in general with other online communities for non-commercial and educational purposes.

Followers