Faster Deployment, Reduced Costs: Optimizing Election Data Analytics on AWS
Industry
DevOps & CyberSecurity
Business Challenge
The client needed to move their infrastructure over to AWS with minimal refactoring and minimal downtime. The business goals were to keep costs down but also allow for bringing new features to market much quicker than the existing model allowed them in a traditional co-location facility. The resulting solution needed to be secure, reliable and have proper monitoring and alerting in place.
Solution
Our solution was to move to a containerized model utilizing several AWS platform services. One issue was that some of the legacy applications were running on SOL Server running on Windows and the ROI for moving completely off this did not make sense. SOL Server on Windows was both costly and inefficient with regards to startup times of servers being deployed and the hourly cost of running those servers. The solution there was to move to SOL Server to Linux. This resulted in cost savings and tremendous speed efficiencies in bringing servers online quickly.
The application architecture was analyzed and where ever possible, refactored to use platform services. This included moving web APls off web servers and over to Lambdas, API Gateway and Dynamo DB. On the container front, AWS ECS and Fargate were used to package up any resources that could not be moved to a serverless, platform services architecture. CloudFormation templates and Jenkins were combined to automate the entire process.
Value Proposition
- Reduced Cost
- Tremendous Efficiencies in Deploying new Version of the Platform
- Efficiencies with Running Multiple Systems at the Same Time
- Reduced Overhead in the Software Development Life Cycle
- Improved auditability of Deployments
Solution Highlights
The simplified isolation of the deployed systems enabled the client to enhance things much faster due to reduced risk in doing so. This was a major benefit to their clients as they were able to bring new features to market and a fraction of the time is took with the previous model.
CLIENT
- Edison Research
INDUSTRY
- News Media and Analytics
REGION
- United States