Well Architected Framework WP - Cost Optimization Flashcards
Cost Optimization Pillar
pay lowest possible price while achieving business objectives
Cost Optimization Design Principles
transparently attribute expenditure
Use managed service to reduce cost of ownership
Swap capital expense for operating expense
Benefit from economies of scale
Stop spending on data center operations
Cost optimization definition - 4 areas
match supply and demand
cost effective resources
expenditure awareness
optimizing over time
Match supply and demand best practice
don’t over or under provision
Use things like autoscaling, or lambda which only responds when there’s a request
Use cloud watch to track demand
2 supply and demand questions
how do make sure capacity matches but doesn’t exceed requirements
how are you optimizing AWS usage
cost effective resources - best practices
using correct instance type is critical.
i.e. run app on T2 micro takes 7 hours and running it on M4.2 Large takes minutes. T2 Micro costs more because took longer
3 cost effective resources questions
Have you selected appropriate resource types for cost targets?
Have you selected appropriate pricing model to meet cost targets?
Are there managed service that will improve ROI
Expenditure Awareness best practices
AWS lets you provision things immediately but that has its own issues.
Be aware of what each team is spending and how
If org has multiple AWS accounts, consolidate billing, use cost allocation tags, billing alerts
Expenditure Awareness questions
what access controls and procedures do you have to govern AWS costs
How do you monitor spending and use
how do you decommission resources, or temporarily stop them, to ensure costs don’t accrue
how do you consider data transfer charges
Optimizing over time
AWS changes frequently. A service chosen yesterday may not be the best one today
Follow the AWS blog and use Trusted Advisor
Optimizing over time questions
how do you mange the adoption of new services
Key AWS Services for these 4 Cost Optimization Areas
Matched supply and demand:
autoscaling
Cost effective resources:
EC2 reserved instances, Trusted Advisor
Expenditure Awareness:
Cloudwatch alarms, SNS, historically consolidated billing and billing alerts
Optimizing over time:
AWS blog, AWS Trusted Advisor
How does AWS recommend optimizing speed for SQS buffer-based approach of matching supply / demand
Add EC2 Spot Instances
How does AWS recommend handling duplicate messages with SQS for buffer-based approach of matching supply / demand
Use FIFO Queues for exactly once processing
How does AWS recommend optimizing Kinesis for buffer-based approach of matching supply / demand
Provision a table in DynamoDB (a NoSQL db)
use idempotent processing so if message processed multiple times, no effect downstream