FinOps on Google Cloud Platform

You have mastered the art of self-healing, doubled down on autoscaling and are ready for success! But all you get at the end of the month is a surprise in the form of a bill several times your budget. Wouldn’t the promise of the cloud that we can achieve the required resilience for our business-critical applications and save money?
In this talk we will offer several cost-saving steps with practical implementation guidelines and examples. We will start with simple, low-hanging fruits and work our way up to a more complicated FinOps setup. The talk focuses on GKE but will touch upon other GCP tools when needed to help you on your way to be the best Sparhamster on GCP.

Transcript

2. 2 HI • Damjan Gjurovski • Platform engineering with focus on GCP
3. HOW TO SAVE MONEY 1. Understanding cloud costs 2. Some low-hanging fruits 3. The toolbox 4.FinOps automation
4. 4 UNDERSTANDING CLOUD COSTS • Usage based - CPU/hour, memory/hour • Data based – Bytes processed; Bytes transferred; Bytes stored • API calls • Fixed prices
5. 5 THE OTHER SIDE • Total cost of ownership • Upfront costs • Developer costs vs managed services • Unexpected demand changes
6. 6 WHAT WE CAN DO TO REDUCE BILLS • Planning • Underutilization • Scaling • Committed use discounts • Preemptable workloads • Managed services/pay-as-you go models • Budget alerts
7. 7 WHAT WE CAN DO TO REDUCE BILLS • Planning • Underutilization • Scaling • Committed use discounts • Preemptable workloads • Managed services/pay-as-you go models • Budget alerts
8. 8 PLANNING • Labels and cost centers • Committed use • Cost estimates
9. 9 PREEMPTABLE WORKLOADS • Workloads that can handle restarts • Workloads that are not time sensitive
10. 10 NODE POOLS • Different node pools for different workloads • Different machine sizes • Scaled down to 0 and waiting for work
11. 11 AUTOSCALERS • Horizontal pod autoscaler • Vertical pod autoscaler • Recommendation mode
12. 12 GKE AUTOPILOT • Pay-per-pod • ”Fixed” service cost
13. 13 THE TOOLBOX • Google Cloud's pricing calculator • Terraform • Infracost • Cloud custodian • FinOps foundation
14. 14 CLOUD PRICING CALCULATOR • Provides an indication for the costs • Might not take into account dynamic scaling • Leads to interesting insights
15. 15 INFRACOST • Can run in the CI pipeline • Provides a cost indication for terraform-managed resources • Can be a part of the MR approval process
16. 16 CLOUD CUSTODIAN • Enforces policies for cloud resources • Filters resources based on type, labels, other properties • Basis for an automated FinOps workflow
17. 17 FINOPS FOUNDATION • FinOps framework • Maturity model • Certifications
18. 18 AUTOMATION • Self-service platforms • Golden path • Enabling engineers to act • Automatically remove unused resources • Cloud scheduler • Cloud functions
19. 19 WANT TO KEEP THE DISCUSSION GOING? MESSAGE ME ON LINKEDIN!