3.8 KiB
title, type, source-type, category, tags, date-added, video-source, audio-source, status
| title | type | source-type | category | tags | date-added | video-source | audio-source | status | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Public Cloud Learning Sessions- Reducing Cloud Costs - 20250318 170100-Meeting Recording | cloud-learning | video | DevOps & SRE/05_FinOps |
|
2026-04-14 | nas:///volume2/work/Public Cloud Learning Sessions/Public Cloud Learning Sessions- Reducing Cloud Costs - 20250318_170100-Meeting Recording.mp4 | summarized (Gemini 摘要) |
Public Cloud Learning Sessions- Reducing Cloud Costs - 20250318 170100-Meeting Recording
Source: NAS /volume2/work/Public Cloud Learning Sessions/Public Cloud Learning Sessions- Reducing Cloud Costs - 20250318_170100-Meeting Recording.mp4
Type: VIDEO | Category: 05_FinOps
Status: 🟡 Awaiting Whisper transcription → Summary
Reducing Cloud Costs
Vinay from the FINOPS team presented a session on reducing cloud costs, focusing on workload and rate optimization. The session covered modernization, right sizing, and best practices for cost reduction.
Workload Optimization via Modernization and Right Sizing
Modernization involves using newer generations of services, like EC2 instances. While there's a perception that newer instances are more expensive, the latest families are generally cheaper and offer better performance. Whenever there's a new family launched by the hyperscale, the latest families are almost cheaper. However, AWS has slightly changed its pricing model after M6, making M7 and M8 somewhat more expensive. Moving from Intel to AMD can save around 6-10% on on-demand prices for Windows and Linux workloads. Graviton instances can offer even greater savings (20-25% reduction in on-demand cost) for Linux workloads, combined with EDP discounts and commitment plans.
Upgrading storage from GP2 to GP3 offers a 20% direct cost benefit without downtime. For Amazon EKS clusters, upgrading to the latest versions is crucial to avoid extended support costs, which are significantly higher. Rather than spending up unnecessary moment on the extended support, you can deploy additional four or five cluster, right. Spot instances can provide up to 90% discount compared to on-demand, suitable for big data, CI/CD pipelines, web servers, and HPC.
Right sizing involves identifying the correct resource configuration for workload performance and capacity needs. The EC2 right sizing recommendation report captures CPU usage, memory, and network data to provide recommendations. Configuring instance schedules is useful for non-production environments, allowing instances to be powered on/off based on business hours, potentially reducing costs to 40% of on-demand prices. Identifying and deleting idle load balancers, unassociated elastic IPs, and underutilized EBS volumes are also key to cost savings. Old snapshots and CloudWatch logs also contribute to unnecessary costs. Using cheaper regions like Oregon or North Virginia can reduce costs if there are no specific regional requirements.
Rate Optimization
Rate optimization involves commitment-based discounts. Hyperscalers offer discounts for committing to resource usage or spending for a term (1-3 years). There are two categories: resource-level commitment (better discount with limitations) and flexible commitment (standard discount with flexibility). AWS offers Savings Plans (EC2 and Compute) and reservations for various services like RDS, ElastiCache, and CloudFront.
The rate optimization workflow includes pre-work (right sizing), analysis (identifying workloads requiring 24/7 uptime), communication (sharing details with finance), approval (from account owner), and reporting (monitoring utilization). Only the Phenop's team can implement commitment plans. All commitment plans will be purchased with no upfront payment options only. The minimum transaction value is 5k per annum.