PROPOSED AWS ARCHITECTURE
About the Customer
Based in Netherlands is a world leading provider of knowledge solutions for the food and beverage industry.
The Challenges
The Application and Database was running on a single DB server and the secondary server was in standby mode. Elasticsearch engine was used for free text search, which was hosted in a single server with limited scalability that made the server crash and had to be rebooted frequently. The application involves image processing tasks, resulting in high CPU and memory utilization. Hosted on a dedicated host, the storage and backups was very expensive.
The Solution
After evaluating available cloud options, we provided AWS solution which was significant with cost optimization and increase in performance. By our analysis, it was identified that 50% of data was not frequently accessed. Hence the database on MS SQL was moved to PostgreSQL and hosted on Amazon RDS. The remaining 50% of frequently accessed database was hosted on MS SQL for high availability.
The Elasticsearch clusters were configured in 18 Amazon EC2 virtual machines, thereby giving room for scalability as and when required.
The image processing tasks was handled through AWS Serverless Architecture concept using the Lambda, SQS & S3 services. AWS Lambda helps in processing millions of images stored in S3 buckets without affecting the application performance.
The user sessions state were processed by 2 nodes of AWS ElasticCache Redis cluster and the load due to all the applications, database and application services was handled through a Load Balancer.
The usage of Amazon S3 along with Amazon Glacier storage for archives reduced the storage cost by 80%. The CloudFront Edge caching helped in content delivery with low latency and high transfer speeds throughout the world.
The Benefits
The database split led to high availability and increase in performance making it easier to handle traffic. Significant reduction in effort to manually set up and upgrade servers due to increase in traffic. The maintenance downtime is reduced with no data loss and 99.99999999999% uptime. The cost savings is about 40% annually when compared to dedicated hosting.