Speaker: James Freeman @ AWS Community Day Hong Kong 2025
Summary by Amazon Nova
https://www.youtube.com/watch?v=ANG8vrtMWDo
On-Premises vs. Cloud Technology Consumption:
On-premises: Procurement cycles, fixed spend, use until hardware fails or contract ends.
Cloud: Pay-as-you-go model, potential for bill shock, cost increases with usage.
Cloud offers low cost of failure, instant procurement, and real-time cost data access.
FinOps Introduction:
FinOps is akin to DevOps, amalgamating finance, technical operations, and technical aspects.
FinOps requires cross-business unit sponsorship and buy-in from executive, finance, procurement, engineering, and operations teams.
Key Considerations for FinOps:
Education is crucial for adopting the FinOps mindset, especially for those transitioning from on-premises models.
Clear accountability and ownership for cloud resource usage and billing are essential.
Emphasizes the need for widespread organizational buy-in for smooth FinOps implementation.
Risk of Losing Control Over Spend:
- Without proper FinOps practices, organizations risk losing control over their cloud spending.
Cost Optimization on AWS:
- Starting resources on AWS is simple, but optimizing for cost requires different considerations.
Importance of Cost Tracking:
Cost tracking is vital for effective FinOps.
AWS uses a "see, save, run" cycle with customers for cost management.
Problem: Visibility for Costs:
- Without visibility into spending and the ability to break down costs meaningfully, meaningful cost management is difficult.
Education and Reviews:
Education about FinOps is key.
Regular reviews are crucial, and AWS TAMs are responsible for organizing these with customers.
Reviews help identify spend patterns and suggest cost-saving options.
Concept of Ownership:
AWS provides constructs to help define ownership and cost accountability.
Using a single AWS account with a corporate credit card for all users is discouraged.
AWS Organizations:
AWS Organizations is a powerful feature for billing, cost management, and security.
It helps in setting up guardrails and managing what users can and can’t do.
It is valuable for new AWS users and has minimal associated cost.
Organizational Units (OUs) and Linked Accounts:
Within AWS Organizations, OUs can be set up similar to directory structures.
Linked accounts can be created for departments or projects to segregate costs and simplify accounting.
Problem: Granular Level of Resources:
At the most detailed level, resources like EC2 instances, Lambdas, and VPCs can become difficult to track.
Organizations may have tens of thousands of these resources, complicating cost management.
Subdivision Within Constructs:
Even with constructs like Organizations, OUs, and linked accounts, further subdivision is often needed.
Cost allocation tagging is used for this purpose.
Cost Allocation Tagging:
Tagging is crucial for cost management, with examples provided (e.g., project, customer, accounts).
There is no right or wrong way to tag, as long as it works for the business.
Emphasis on creating a tagging dictionary to avoid confusion and ensure consistency.
Example: Differentiating between "Project," "project," "project-name," and "project_name" due to case sensitivity and varying formats.
Tagging Dictionary:
A tagging dictionary is simple but vital to prevent mix-ups and ensure clear cost allocation.
Helps decipher spending by project or cost center.
Cost Visibility:
- At the end of the month, AWS sends a bill with a total number, which is a starting point for cost visibility.
Ensuring Cost Control:
Importance of oversight to determine if spending is appropriate and to identify potential shadow IT projects.
Questions to ask and monitor regarding AWS spend.
AWS Cost Explorer:
A free tool similar to Excel graphing, allowing detailed filtering and data diving.
Highly recommended for its power and ease of use.
AWS Cost and Usage Reports:
For more detailed data than Cost Explorer, with a small cost associated for storing data in S3 and using tools like Athena.
Useful for complex setups and specific internal accounting metrics (e.g., vCPU hours).
Cloud Intelligence Dashboards:
Free, open-source dashboards that cut data in various ways and present it digestibly.
Allow breakdown of instance, storage, and networking spend.
Available on GitHub for deployment, with cost mainly being the deployment in Quicksight.
Dashboards range from high-level (suitable for executives and finance) to granular (for fin practitioners and engineering).
Value depends on the user's role within the business.
https://docs.aws.amazon.com/guidance/latest/cloud-intelligence-dashboards/dashboards.html
Awareness and Usage:
The goal is to raise awareness of these tools and their suitability for different business levels.
Significant engineering effort has been put into creating these dashboards for appropriate use cases.
Reducing AWS Bills and Saving Money:
- The presentation covers a runbook for cost-saving strategies, focusing on cost savings impact (y-axis) and technical complexity (x-axis).
Simple Cost-Saving Measures:
Commitments: Using instance savings plans and reserved instances for resources running 365 days a year can provide significant discounts on on-demand rates.
Elastic Workloads: Shutting down resources during periods of low usage (e.g., weekends) can lead to immediate cost savings.
Identifying Underutilized Resources: Tools like Cost Explorer and Cloud Intelligence Dashboards help identify and delete idle resources.
Moderate Complexity Measures:
Serverless Architecture: Moving to serverless platforms (e.g., RDS) can reduce costs by paying only for compute time needed.
Graviton Instances: Transitioning from x86 to ARM 64 architectures (Graviton) offers significant cost savings with no impact on software.
EBS Volumes: Upgrading from GP2 to GP3 EBS volumes can save money.
High Complexity Measures:
Right Sizing: Optimizing instance types and resources to match actual usage.
Cloud-Native Architecture: Moving away from static resources (e.g., EC2, RDS) to fully serverless, pay-per-use models.
Long-Term Strategy: Adopting cloud-native architectures is the most powerful lever for long-term cost savings, though it requires significant engineering and development effort.
Unit Cost as a Metric:
Measuring unit cost (e.g., cost per transaction) rather than overall spend.
Decreasing unit cost over time as a sign of efficiency.
Key Performance Indicators (KPIs):
Suggest picking three KPIs tailored to business requirements.
Examples: unit cost, customer satisfaction, and efficiency metrics.
Importance of defining and understanding unit cost for optimization.
Feedback Loops:
Continuous feedback loops for ongoing optimization.
Feedback should be a regular, iterative process, not a one-time event.
Prioritization and Communication:
Businesses should determine what to prioritize based on customer satisfaction and efficiency.
Effective communication is crucial for complete feedback loops.
Ensure open conversations between finance, procurement, and engineering teams.
Concerns and insights should be communicated and integrated into the roadmap.
Incomplete loops (lack of communication) indicate problems in the process.
Reviewing with Cloud Intelligence Dashboards:
Use dashboards for customer reviews, tailoring data to stakeholders’ needs.
C-level executives require top-level data and KPIs, not detailed instance breakdowns.
Consider the audience and format when communicating data.
Row-Level Security:
Out-of-the-box dashboards provide universal access; use row-level security for granular control.
Restrict access to relevant data for specific departments to maintain security and relevance.
Education on Cost Management Tools:
Emphasize the importance of educating teams responsible for cost, especially engineers.
Tools like Cost Explorer, budget settings, commitment recommendations, anomaly detection, and right-sizing recommendations are available.
Anomaly detection and right-sizing recommendations help optimize resource usage.
Cost and Usage Reports provide detailed data, though they incur a small cost.
Additional AWS Tools for Cost Optimization:
Compute Optimizer: A free tool worth exploring to further optimize EC2 instance costs.
Trusted Advisor: Provides numerous recommendations, not just for cost optimization but also for security.
AWS Config: Often needed by customers for various purposes.
CloudWatch: Critical for gathering metrics within EC2 instances, as AWS cannot see into customer instances by default.
S3 Lens: A powerful tool for visualizing and optimizing the use of S3 object storage.
Shared Responsibility Model:
- AWS is responsible for the integrity of the cloud, while customers are responsible for what they run on the cloud.
Cost of Chargeable Services:
Tools like CloudWatch are essential for customers to gather metrics and optimize their resources, as AWS cannot access this data directly.
Even chargeable services like Cost and Usage Reports have negligible costs, primarily due to S3 data storage.
These tools are valuable for detailed cost optimization and should be considered despite the small associated costs.
Team:
AWS FSI Customer Acceleration Hong Kong
AWS Amarathon Fan Club
AWS Community Builder Hong Kong
Top comments (0)