Cost Optimizations In Cloud The best way to run a continuous optimization cycle

Cost Optimizations In Cloud
The best way to run a continuous optimization cycle

1. Cost Optimization in Cloud : The best way to run a continuous optimization cycle 202 4 / 08 /2 4 アイレット株式会社 檜垣 慶太 コスト最適化持続的サイクルの最善の方法

2. Who AM I ‹#› Name ：Keita Higaki (X : @keitah0322 ） Belongs ：Iret,Inc Division ： Customers Support Division Role ：Technical Account Manager / Gloup Leader 　　　 Prize ：2023 Japan AWS Top Engineers 　　 ：2024 Japan AWS Top Engineers ：Google Partner Top Engineer 2024 Speciality ：Serverless Architechture

3. First of All ‹#›

4. ‹#› Potential to reduce cloud usage fees It is said that customers who are already using the cloud have the potential to reduce costs by an average of 30%！ (*1). (*1) 出典 : 2021 Flexera State of the Cloud Report, N=750 70% Efficient cloud spending 30% Excessive spending ‹#›

5. ‹#› So how can you optimize costs?

6. ‹#› What is the continuous cost optimization cycle? With the goal of helping customers reduce costs on an ongoing basis, we visualize their cloud environment, analyze their cost structure, evaluate cost savings, and then help them plan and forecast future expenditures. Through these efforts, we strengthen financial management of cloud usage and even draft a three-year cloud usage plan. Visualization Planning and Forecasting optimization クイック ウィン 最適化 アーキテクチャ最適化 FinOps in Action Realization of a continuous optimization cycle クラウド使用量予測 予算策定

7. Step 1 Cost visibility is a crucial first step in the optimization process！ ‹#›

8. Step 1 Visualize 　~Dashbord~ ‹#› To visualize AWS costs using Amazon QuickSigh t via the AWS Cost & Utilization Report (CUR). Cost visibility is a crucial first step in the optimization process！

9. Step 1 　Visualize　〜Reporting〜 ‹#› Reporting including evaluation of reduction potential 　

10. Step2 Introduction to optimization operations ‹#›

11. Reference 　Two techniques for cost optimization ‹#› Cost optimization techniques can be broadly divided into two categories : quick-win optimization and architecture optimization for cloud-native deployment. Quick Win optimization Scheduling according to demand Reduce the scale an appropriate instance Use Cloud Native ‹#› １ ２ ３ ４ 5 Re-Architecture shutting down unused resources Reselecting the right storage

12. Step2 　 Quick Win Optimization ‹#› Reselect instances (size, family, generation/AMD/Graviton, region) and purchase options based on resource utilization and workload characteristics m4.4xlarge (Linux) m4.xlarge (Linux) Sizing/Family Changes Change to latest generation /AMD/Graviton Region Change Reselect instance 　　　　　　　 Change purchase option Solution Change to On-Demand, RI/Savings Plans, or Spot depending on operating hours RI/ Savings Plans Spot Tokyo t2.large (Linux) $0.1216/h t3a.large (Linux) $0.0979/h m5.large (Linux) $0.124/h m5.large (Linux) $0.096/h Virginia ‹#› Use Cost Explorer To Identify Ri/Savings Plans Recommendations on demand Use Compute Optimizer To make sure Optimal Instance and EBS for Our Workloads

13. Step2　Use resources according to Workload ‹#› 25 従来の使用容量 容量 ニーズ ニーズに応じたリソース適応 時間 Scheduling by deploying AWS Instance Scheduler We want to reduce unnecessary operation by stopping instances when they are not being used and adaptively adjusting resources according to demand. In fact it's very tough here. 稼働日 ： 稼働インスタンス ： 需要 未使用リソース削減 Automatically increase or decrease instances using Auto Scaling

14. How to delete unused resources ‹#› Use AWS Trusted Advisor to identify unused resources. Remove unused resources using Trusted Advisor Identifying idle EC2 instances using Cost Explorer NW Unattached Elastic IP Idle/Unattached Load Balancer Storage Unattached EBS Instance Idle EC2, RDS, Redshift Extracting unused resources Solution ‹#›

15. Step2　Purchasing Strategy ‹#› Savings Plans/RIs are used for workloads with fixed usage/production/stable demand. 1/3 year contract commitment, saving up to 72% compared to on-demand On-Demand is used for new/non-production/hard to predict workloads Billed by the second It is important t o choose the right combination of purchasing options based on the characteristics of your workload. By utilizing unused AWS resources, you can save up to 90% compared to on-demand. ‹#› Spot is used for fault-tolerant, flexible, and stateless workloads.

16. Step2　Purchasing Strategy ‹#› Reference: Typical purchase option patterns Spot On demand SP/RI ‹#› Data analysis Internal IT New application development Development/Test Environment Spot On demand SP/RI Spot On demand SP/RI Spot On demand SP/RI

17. Architecture Optimization ‹#›

18. Architecture Optimization Strategy ‹#› Public subnet Private subnet Private subnet Virtual private cloud (VPC) Dynamodb API Gateway Using serverless/managed services improves maintainability and enables you to focus on business logic, which is expected to speed up development. However, there is a trade-off as replacement requires effort .

19. Resource-specific Replace strategies ‹#› Replace serverless solutions such as Aurora Serverless and Dynamo DB Containerize or replace with serverless solutions such as Lambda There are plenty of storage optimization solutions available. ElastiCache Serverless

20. Using 3rd Party Solution via MarketPlace ‹#› We can expand our selection via MarketPlace. Sometimes a third party may be more appropriate. It's important to have many options and be flexible.

21. Step3 Plan & Forecast ‹#›

22. Step3　Planning and Forecasting Methodology ‹#› It is essential to plan and forecast cloud usage , implement unit economics, and track sustained cost optimization. 現時点 単価 (cost/transaction) 利用料金 時間 Solution Machine learning pricing forecasting with Amazon Forecast Forecasting with AWS Cost Explorer Set budgets and forecast usage costs with AWS Budgets 将来予測 ‹#› Usage fee forecast image

23. How to achieve sustainable cost optimization ‹#›

24. The need for a dedicated cost analysis team ‹#› In order to advance cloud native, IT and finance teams need to come together. SCM Ops Organizational image in the on-premise era Organizational image in the cloud era Separation of IT and finance allows for a long time between procurement and deployment to production By integrating IT and finance, we can quickly deploy and operate the system and optimize it. Approve Supply Construct Demand IT PUR Ops Optimizations Deploy IT Finance

25. Road to FinOps ! ‹#› It is necessary to build processes for each area that requires individual optimization based on workload , and for areas that should be targeted for overall optimization such as RI/SP . Overall optimi.zation through CCoE Individual optimization by project teams, Process Setting shared resource allocation rules / Setting and managing KPIs common to all users Cost management of shared resources Define and measure KPIs that incorporate the value provided by the system into infrastructure costs Account splitting / tagging Account division/tagging rules established Volume discounts through account consolidation Account ManagementTagging Instance Selection Standardization of instance types and services used Optimizing the system by adjusting instance size and operation purchase options Efficient use of discount programs through RI/SP sharing Individual RI/SP purchases Architechture Reference architecture development Architecture Optimization

26. Summary ‹#› Cost visualization Case study organization for cloud optimization, Action！ Build automation into the architecture and adjust scheduling！ Manage AWS budgets by organization and execute planning and forecasting Formation of a cross-functional organization between the finance and IT departments to achieve sustainable cost optimization Let's enjoy the wonderful cloud-native world!

27. ‹#› Thank you for listening.