| Table of Contents |
|---|
Quickstart
https://obrienlabs.medium.com/cloud-finops-18a5e9942d84?sk=88b1852b11d4dfd6e8c84cca9fd83350
We watch the AWS Cost Explorer https://console.aws.amazon.com/cost-management/home?#/dashboard to manage our on-demand, reserved and spot costs.
Cloud Financing
Cloud Free Tier
| Google Cloud | https://cloud.google.com/free/docs/gcp-free-tier#free-tier-usage-limits |
| AWS | sort on "Always free" |
| Azure | |
| IBM Cloud | https://www.ibm.com/cloud/free |
| Oracle Cloud |
Cloud Startup Programs
AWS Activate Founders
https://aws.amazon.com/activate/founders/
AWS Activate will require a project that is running 100% on amazon - you will receive the default $1000US one time funds - if you meet the criteria you can move up to 100K
Cloud Training
AWS Partner status
https://partnercentral.awspartner.com/APNLogin
https://www.exitcertified.com/it-training/aws/architect/architecting-4-0-34767-detail.html
AWS Partner status has additional partner courses and 75% off select courses above at Exit Certified - even while you are in the "registered" status - before you meet all criteria
AWS Customer Counsel
Assist in customer insight
$50 for surveys
$150 for meetings (I missed the one on the 2nd of Aug as I broke my shoulder with a quick last minute rollerblade 2h before the meet - first time in 30 years I went without foot inserts as I forgot them in my other rollerblades after the rivet on the 2021 models broke on my 2nd pair - hence the need lately to purchase 4+ blade pairs in advance from K2)
FinOps Principals
Favour Autoscaling to follow the demand curve very closely
...
Autoscaling is not instant. The underlying infrastructure of NLB autoscaling for example - is itself EC2 instances - that take time to replicate and start. Even Lambda needs to be pre-warmed. Therefore the disadvantage to in placing some of a K8S cluster worker nodes under an auto scaler for example will be that that capacity will not be instantly available like it would if we overprovisioned over provisioned for excess capacity ahead of time.
...
| Type | Granularity | Service | Example | Utilization per service | Type | Formula | Free Tier | Cost US/m | Shared overhead |
|---|---|---|---|---|---|---|---|---|---|
| compute | 1 vCPU | IaaS EC2 | t3a.micro | 100% | On Demand | ||||
| Reserved 3y no front | |||||||||
| Spot 20210218 | |||||||||
| PaaS K8S | 3 x t3a.large | 1/12 | |||||||
| CaaS Fargate | n/a | ||||||||
| FaaS Lambda | n/a | 1M 128Kb 100ms = .0125 GB-s | 400k GB-s | 0.20 req 0.21 exec $0.41 | |||||
| persistence | 1 GB | IaaS RDS | 100% | ||||||
| DBaaS Aurora | n/a | ||||||||
| DBaaS DynamoDB | n/a | ||||||||
| storage | 1 GB | S3 | n/a | ||||||
| throughput | 1 Gbps | Network In | |||||||
| AI AWS Textract | Text and Image Processing#TextractAPIExamples | 0.07 / tx
|
Costing Options
| Tool | Details | |
|---|---|---|
| Cost Explorer | ||
| Cost Estimator | ||
| Savings Plans | ||
| Free tier usage | Most services have tier - once used gone - so the first service in gets the benefit | |
| Volume pricing | If several services saturate for example S3 - subsequent services will benefit with lower pricing (resource pooling) | |
| Auto scaled reserved | If service A kicks in k8s autoscaling of the worker nodes - all other services benefit by default due the capacity increase. The reverse is true - if service A terminates - then service B (rogue) had full use of most of the vCores on a scaled node - now needs to share in a more overall saturated smaller cluster | |
| Partitioned use | Move read-only traffic - like monitoring/reporting to a read replica that is optimized for read not read/write |
...
Presentation
| Label/Tag | Details | Examples |
|---|---|---|
| intro | several layers of costs savings | |
| History | discuss how aws sold excess monthly capacity detail spot | |
| Effects | cost model can drive innovation cloud adoption is always hybrid | |
| Levels | reserved preemptible - google fixed spot 24h spot - 2 min warning kubernetes PaaS serverless "cloud is the new mainframe" | Amazon EC2 Spot - show savings on wizard |
| serverless | ||
| demo | ec2 reserved ec2 spot rds reserved lambda resource limits | |
| Most orgs not in the business of IT | ||
| drawbacks | serverless issues of infinite scale resource limits not set | Kubernetes Cluster across VMware nodes on OSX or Windows#Experiment:RunafullsaturationDaemonSetkubernetesdeploymentacrossallnodesinthecluster |
| security by obscurity - individual EC2 hacking | ||
| best practices | immutable infrastructure automated deployment/scaling |
Links
https://www.finops.org/events/
Form3 Cloud Native Payments (Startup) https://www.form3.tech/about