Non-Disruptive Migration of a 20-Server Ecommerce Platform to AWS
About Our Client
Headquartered in the United States, the Client is a leading provider of innovative online shopping, travel and entertainment solutions to consumers in more than 100 countries. The Client’s main focus is to offer cashback rewards for each purchase on their website featuring millions of products and services sold by the most popular online stores along with the leading travel, telecommunications and financial services companies from around the globe.
Challenge
The Client had several services (a web portal, statistics, payments, commission, mailings and others) running on 20 dedicated servers. With the ever-rising number of the website visitors, the high server loads began to cause poor services performance. This resulted in a significant traffic decline followed by the drop in revenue as fewer consumers could make purchases on the website. The challenge was to migrate the entire infrastructure from the dedicated servers to AWS (Amazon Web Services), thus ensuring scalability of the Client’s services.
Solution
ScienceSoft's development team comprised of a PHP Architect and a Senior PHP Developer started off with migrating the Client’s six database servers which totaled 100GB. Using the master-slave scheme, they managed to replicate the data to AWS without any downtime. The web-servers were migrated to Amazon EC2 (Elastic Compute Cloud) and configured based on Red Hat Linux.
The next step was to modify the Client’s services, so that they follow a shared-nothing pattern, a concept which stands for a system where each node has its own memory and disk storage. The idea was to distribute the data used by one application among a number of external storages, such as Amazon S3 (Simple Storage Service), RDS (Rational Database Service) and DynamoDB (NoSQL database service), to make the application more scalable.
For a better control over the virtual networking environment, ScienceSoft applied Amazon VPC (Virtual Private Cloud) to create public and private subnets and establish multiple levels of security for the Client’s services.
To answer the Client’s backup needs, ScienceSoft developers wrote scripts that would regularly archive the data and create database snapshots.
Results
The migration was successfully completed in 1.5 months and brought about a cloud-operating, scalable production system able to withstand high loads. The Client can now avoid the considerable expenses caused by running their services on dedicated servers. Together with ScienceSoft, the Client is also pondering the idea of consolidating all the databases to facilitate data partitioning and ensure better scalability.
Technologies and Tools
Amazon Web Services: VPC, EC2, S3, RDS, Route 53, CloudFront CDN, DynamoDB
