Cost optimization review
The Well-Architected Assessment is a top-down review of your workload to identify what could be improved in your architecture. In this step we will start focussing on the costs of your workload, what are the most common pain points and how to improve them. We will be doing this work in three levels, so you can go deeper if you want or just scratch the surface when you need to focus on other pillars first. This will help on the most immediate and direct impact of the recommendations.
Level 1 - Cost Management, Sizing and Reservations
When the solution is already deployed in Azure, you first stop is the Azure Advisor where you will find recommendations based on the telemetry of your workload. With this telemetry you already have the usage patterns and a lot more information about the workload. This information helps on building meaningful recommendations.
Take a look to the cost over the past 12 months to identify outliners, and compare it to the budget you have for this application.
Check that there are alerts configured and that they are aligned with the budget.
Get an estimation of the operational costs of the current architecture.
Review that the size and SKUs of the services is aligned with the utilization patterns, and check auto-shutdown, scaling and using spot instances where available.
Using the 12 month costs, check if you can apply reservations on your resources, know what type of resources you can reserve (compute, database, storage, caching…). Find if you can apply to other discounts like the hybrid benefits and Dev/Test licenses.
Summary
- Look into Azure Advisor
- Get the past 12 month costs
- Review the budget
- Check alerts
- Calculate operational costs
- Check reservations and other offers
Level 2 - Resources and Locations
Review unused resources and if you have the mechanisms to stop them or having environments on demand. Review the retention time for backups, and check that you backup only the data that cannot be recovered by other means like Infrastructure as Code. Check your resources location, some regions are cheaper, but take into account egress, peering and gateway costs. Analyze disk performances and replication configuration, for example: what are your replication needs for the storage accounts?
Level 3 - Architecture changes
There are architecture patterns that can reduce costs. Check the use of workers, queues, caching and sharding. Is it possible to extract the most intensive processes to on-demand workers? This would reduce the base cost of the solution by reducing the SKUs where the main workload is deployed.
Check the use of compression and code optimizations.
Evaluate the migration to PaaS or Container services to enhance costs.
Resources
Checkpoint
- A clear Cost Model for the workload
- Cost savings calculations for the most important recommendations
- Effort evaluation for optimizations related to cost
- Cost trade-offs
Now you can move to the next pillar: Security