SaaS Product Cloud Refactor
Jan 2019 – Sep 2019 · PRMconnect
- Microsoft Azure
- PaaS
- Cloud Architecture
Moved a global SaaS product from a legacy data center to Azure PaaS, reaching average uptime above 99.99%, with total cost after 1 year 68% lower than the data center budget.
Challenge
A global SaaS provider was experiencing scaling and reliability issues in a legacy data center, where average uptime for the core product had fallen below 99%. Two upcoming large customer projects were expected to exceed existing capacity, potentially causing slow response times and errors across multiple end customers. Azure was chosen as the target cloud provider, with the goal of using PaaS services as much as feasible and beating the budget of a traditional data center build.
Analysis
- Profiled applications on the existing architecture to understand performance requirements
- Audited each aspect of the architecture to determine fit with PaaS offerings
- Identified dependencies that were not immediately compatible with cloud architecture
- Analyzed application code for direct references to the file system, operating system and network
- Wrote a budget for all required Azure resources, for comparison against the cost of a physical data center build
Approach
- Developed a plan that mapped current resources to cloud resources and highlighted blocking elements in the current architecture
- Defined scopes to remediate all blocked systems and interfaces, and discussed the cost-benefit of each with the customer
- Provisioned scalable PaaS architecture in Azure
- Implemented a pilot program to run upcoming large-scale events on the new infrastructure
Results
- The pilot program was implemented on time, with all performance benchmarks exceeded
- Within 6 months of the pilot, all customers had been transitioned to the cloud environment
- Average uptime above 99.9% was achieved, with all downtime attributable to failures in Azure PaaS
- High-availability features were later improved to achieve average uptime above 99.99%
- After 1 year, total cost was 68% lower than the traditional data center budget, with a projected 38% total cost reduction over the expected life of a traditional data center