feat: cost-saving measures

knowledge base/cloud computing/aws/cloudwatch.md

@@ -5,6 +5,7 @@ Observability service. with functions for logging, monitoring and alerting.
 1. [TL;DR](#tldr)
 1. [Queries of interest](#queries-of-interest)
 1. [Stream logs](#stream-logs)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
    1. [Sources](#sources)
 
@@ -31,6 +32,19 @@ The [CloudWatch console] offers some default good queries.
 
 Logs in Log Groups can be [streamed][stream logs] elsewhere.
 
+CloudWatch retains metrics' data as follows:
+
+- Data points with a period of less than 60 seconds are available for 3 hours.<br/>
+  These are high-resolution custom metrics.
+- Data points with a period of 60 seconds (1 minute) are available for 15 days.
+- Data points with a period of 300 seconds (5 minutes) are available for 63 days.
+- Data points with a period of 3600 seconds (1 hour) are available for 455 days (15 months).
+
+Data points are aggregated together for long-term storage after the initial period.<br/>
+E.g., data using a period of 1 minute remains available for 15 days with 1-minute resolution, then it is aggregated and
+made available with a resolution of 5 minutes; after 63 days, it is further aggregated and made available with a
+resolution of 1 hour for 15 months.
+
 <details>
   <summary>CLI commands</summary>
 
@@ -101,6 +115,12 @@ Also refer [Streaming CloudWatch Logs data to Amazon OpenSearch Service] to stre
 
 Logs in CloudWatch Log Groups can be streamed [Kinesis], [Firehose] or [Lambda] by leveraging Logs subscriptions.
 
+## Cost-saving measures
+
+- Configure an _appropriate_ log retention period for any log groups.<br/>
+  Log groups containing development logs should not usually need more than 1w worth.
+- When in doubt, still configure a default, long log retention period for all log groups (10y?).
+
 ## Further readings
 
 - [Website]
@@ -113,6 +133,7 @@ Logs in CloudWatch Log Groups can be streamed [Kinesis], [Firehose] or [Lambda]
 - [Real-time processing of log data with subscriptions]
 - [Streaming CloudWatch Logs data to Amazon OpenSearch Service]
 - [Which log group is causing a sudden increase in my CloudWatch Logs bill?]
+- [Metrics concepts]
 
 <!--
   Reference
@@ -130,6 +151,7 @@ Logs in CloudWatch Log Groups can be streamed [Kinesis], [Firehose] or [Lambda]
 [firehose]: https://docs.aws.amazon.com/firehose/latest/dev/what-is-this-service.html
 [kinesis]: https://docs.aws.amazon.com/kinesis/
 [lambda]: https://docs.aws.amazon.com/lambda/
+[Metrics concepts]: https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/cloudwatch_concepts.html
 [real-time processing of log data with subscriptions]: https://docs.aws.amazon.com/cloudwatch/latest/logs/Subscriptions.html
 [services that publish cloudwatch metrics]: https://docs.aws.amazon.com/cloudwatch/latest/monitoring/aws-services-cloudwatch-metrics.html
 [streaming cloudwatch logs data to amazon opensearch service]: https://docs.aws.amazon.com/cloudwatch/latest/logs/CWL_OpenSearch_Stream.html

knowledge base/cloud computing/aws/ebs.md

@@ -6,9 +6,11 @@ Persistent [block storage][what is block storage?] for [EC2 Instances][ec2].
 1. [Volume types](#volume-types)
 1. [Snapshots](#snapshots)
 1. [Encryption](#encryption)
+1. [Archiving](#archiving)
 1. [Operations](#operations)
    1. [Increase disks' size](#increase-disks-size)
    1. [Migrate `gp2` volumes to `gp3`](#migrate-gp2-volumes-to-gp3)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
    1. [Sources](#sources)
 
@@ -79,14 +81,16 @@ details about EBS balances.
 
 ## Volume types
 
-Refer [Amazon EBS volume types].
+Refer [Amazon EBS volume types] and [Amazon EBS General Purpose SSD volumes].
 
 |                     | `gp3`                                            | `gp2`          | `io2`             | `io1`             | `st1`            | `sc1`            |
 | ------------------- | ------------------------------------------------ | -------------- | ----------------- | ----------------- | ---------------- | ---------------- |
 | Class               | SSD                                              | SSD            | SSD               | SSD               | HDD              | HDD              |
 | Annual failure rate | 0.1% - 0.2%                                      | 0.1% - 0.2%    | 0.001%            | 0.1% - 0.2%       | 0.1% - 0.2%      | 0.1% - 0.2%      |
 | Size                | 1 GiB - 16 TiB                                   | 1 GiB - 16 TiB | 4 GiB - 64 TiB    | 4 GiB - 16 TiB    | 125 GiB - 16 TiB | 125 GiB - 16 TiB |
+| Baseline IOPS       | 3000                                             | 100            | 4,000             | 100               | N/A              | N/A              |
 | Max IOPS            | 16,000                                           | 16,000         | 256,000           | 64,000            | 500              | 250              |
+| Baseline throughput | 125 MiB/s                                        | 128 MiB/s      | 1,000 MiB/s       | 1 MiB/s           | 5 MiB/s          | 1.5 MiB/s        |
 | Max throughput      | 1,000 MiB/s                                      | 250 MiB/s      | 4,000 MiB/s       | 1,000 MiB/s       | 500 MiB/s        | 250 MiB/s        |
 | Multi-attach        | No                                               | No             | Yes               | Yes               | No               | No               |
 | NVMe reservations   | No                                               | No             | Yes               | No                | No               | No               |
@@ -117,6 +121,33 @@ Total: $1.71 + $0.00 + $1.66 = $3.37
 
 </details>
 
+`gp3` volumes are normally much better and more cost-effective than `gp2` ones.<br/>
+There are still specific situations where `gp2` volumes _might_ be _slightly_ better than `gp3` ones, namely:
+
+- Burst-based, spiky workloads with low sustained demand, like cron jobs, backups, compactions, and batch analytics.
+
+  <details>
+
+  `gp2` volumes accumulate burst credits when they are underutilized.<br/>
+  Large volumes accumulate credits quickly, and can then burst for a long time.
+
+  Sustained performance for `gp3` volumes costs a pretty penny, even if it's only needed briefly.
+
+  </details>
+
+- Very large volumes (> 1 TiB) that do not require guaranteed throughput tuning.
+
+  <details>
+
+  `gp2` volumes' performance is based on their size, and larger volumes can provide more **baseline** performance for
+  free.<br/>
+  E.g., at 4 TB a `gp2` volume as a baseline of 12000 IOPS, while a `gp3` volume still has a baseline of 3000 IOPS.
+
+  The maximum throughput will still be lower than `gp3` volumes, but as long as up to 250 MiB/s in bursts is fine it
+  can be a better deal.
+
+  </details>
+
 ## Snapshots
 
 A volume's first snapshot is a **complete** snapshot of it, with _all the volume's blocks_ being copied over.<br/>
@@ -218,6 +249,12 @@ Attaching EBS volumes which data keys are encrypted with unusable KMS keys to EC
 not be able to use the KMS keys to decrypt the data key used for the volume.<br/>
 Make the KMS key usable again to be able to attach such EBS volumes.
 
+## Archiving
+
+Refer [Amazon EBS General Purpose SSD volumes].
+
+Archiving has a 90d minimum storage fee, **and** archived resources have retrieval fees.
+
 ## Operations
 
 ### Increase disks' size
@@ -279,6 +316,16 @@ If changing the volume type from `gp2` to `gp3` **without** specifying IOPS or t
 automatically provisions either equivalent performance to that of the source `gp2` volume, or the baseline `gp3`
 performance, whichever is higher.
 
+## Cost-saving measures
+
+- Prefer using `gp3` volumes unless an application requires specific IOPS or throughput.
+- Still prefer `gp3` volumes to `gp2`.<br/>
+  `gp3` volumes cost less, and have better performance per GB (except some specific corner cases).<br/>
+  Performance of `gp3` volumes can also be somewhat tuned, while `gp2`'s only increase with size.
+- Consider using `gp2` volumes _only_ when encountering those corner cases, usually where size > 1 TiB and comparable
+  higher-than-baseline bandwidth is needed only in bursts.
+- Consider [archiving] when snapshots should not be accessed for 90d or more.
+
 ## Further readings
 
 - [Amazon Web Services]
@@ -309,16 +356,20 @@ performance, whichever is higher.
   ═╬═Time══
   -->
 
+<!-- In-article sections -->
+[archiving]: #archiving
+
 <!-- Knowledge base -->
 [amazon web services]: README.md
 [cli]: cli.md
 [ec2]: ec2.md
 
 <!-- Upstream -->
+[Amazon EBS General Purpose SSD volumes]: https://docs.aws.amazon.com/ebs/latest/userguide/general-purpose.html
 [amazon ebs pricing]: https://aws.amazon.com/ebs/pricing/
 [amazon ebs volume types]: https://docs.aws.amazon.com/ebs/latest/userguide/ebs-volume-types.html
 [amazon ebs-optimized instance types]: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-optimized.html
-[archive amazon ebs snapshots]: https://docs.aws.amazon.com/ebs/latest/userguide/snapshot-archive.html
+[Archive Amazon EBS snapshots]: https://docs.aws.amazon.com/ebs/latest/userguide/snapshot-archive.html
 [automate snapshot lifecycles]: https://docs.aws.amazon.com/ebs/latest/userguide/snapshot-ami-policy.html
 [choose the best amazon ebs volume type for your self-managed database deployment]: https://aws.amazon.com/blogs/storage/how-to-choose-the-best-amazon-ebs-volume-type-for-your-self-managed-database-deployment/
 [delete-volume]: https://docs.aws.amazon.com/cli/latest/reference/ec2/delete-volume.html

knowledge base/cloud computing/aws/ec2.md

@@ -9,6 +9,7 @@
    1. [Lifecycle hooks](#lifecycle-hooks)
 1. [Image customization](#image-customization)
 1. [Automatic recovery](#automatic-recovery)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
    1. [Sources](#sources)
 
@@ -19,13 +20,15 @@ The API for EC2 are [**eventually** consistent][Eventual consistency in the Amaz
 EC2 instances are billed by the second, with a minimum of 60s,
 [since 2017-10-02][announcing amazon ec2 per second billing].
 
-Use an instance profile to allow an EC2 instance to use an IAM role.
+Use an IAM Instance Profile to allow an EC2 instance to use an IAM role.
 
 `T` instances launch as `unlimited` by default. Launch them in `standard` mode to avoid paying for surplus credits.
 
 The instance type [_can_ be changed][change the instance type]. The procedure depends on the root volume, and **does**
 require downtime.
 
+When using spot instances, prefer instrumenting the application to be aware of [termination notifications].
+
 Clone EC2 instances by:
 
 1. Creating an AMI from the original instance.
@@ -222,6 +225,25 @@ Refer [Image Builder].
 
 Also see [Automatic instance recovery].
 
+## Cost-saving measures
+
+- Prefer using the most adequate instance type for the job.<br/>
+  E.g., prefer `r*` instances instead of `m*` ones where a lot of RAM is needed, but almost no CPU power is.
+- Prefer using ARM-based (`g`) instances, unless a different architecture is required.
+- Prefer _shared_ instances over _dedicated_ ones unless necessary.
+  Refer [Understanding AWS Tenancy Options].
+- Prefer dedicated _instances_ over dedicated _hosts_ unless necessary.
+  Refer [Understanding AWS Tenancy Options].
+- Prefer using [burstable (`t`) instances][burstable instances], unless steady performance is required and specially
+  for burstable workloads.
+- When employing **underused** burstable instances, prefer re-launching them in `standard` mode to avoid paying for
+  surplus credits.
+- Prefer using [spot instances] instead of on-demand ones where possible.
+- Consider **stopping** or (even better) deleting non-production hosts after working hours.
+- Consider applying for EC2 Instance and/or Compute Savings Plans.
+- Consider [archiving snapshots] should they not be accessed for 90d or more.<br/>
+  Archiving has a 90d minimum storage fee, **and** archived resources have retrieval fees.
+
 ## Further readings
 
 - [Amazon Web Services]
@@ -269,8 +291,12 @@ Also see [Automatic instance recovery].
   ═╬═Time══
   -->
 
+<!-- In-article sections -->
+[burstable instances]: #burstable-instances
+
 <!-- Knowledge base -->
 [amazon web services]: README.md
+[archiving snapshots]: ebs.md#archiving
 [cli]: cli.md
 [ebs]: ebs.md
 [image builder]: image%20builder.md
@@ -302,7 +328,9 @@ Also see [Automatic instance recovery].
 [Manually create or edit the CloudWatch agent configuration file]: https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/CloudWatch-Agent-Configuration-File-Details.html
 [recommended alarms]: https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/Best_Practice_Recommended_Alarms_AWS_Services.html#EC2
 [retrieve instance metadata]: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/instancedata-data-retrieval.html
+[Spot Instances]: https://aws.amazon.com/ec2/spot/
 [standard mode for burstable performance instances]: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-performance-instances-standard-mode.html
+[termination notifications]: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/spot-instance-termination-notices.html
 [unlimited mode for burstable performance instances]: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-performance-instances-unlimited-mode.html
 [using al2023 based amazon ecs amis to host containerized workloads]: https://docs.aws.amazon.com/linux/al2023/ug/ecs.html
 [using instance profiles]: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_use_switch-role-ec2_instance-profiles.html

knowledge base/cloud computing/aws/ecs.md

@@ -36,6 +36,7 @@
 1. [Best practices](#best-practices)
 1. [Troubleshooting](#troubleshooting)
     1. [Invalid 'cpu' setting for task](#invalid-cpu-setting-for-task)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
     1. [Sources](#sources)
 
@@ -133,6 +134,12 @@ curl -fs "http://$( \
   )" --query "tasks[].attachments[].details[?(name=='privateDnsName')].value" --output 'text' \
 ):8080"
 
+# Get the image of specific containers.
+aws ecs list-tasks --cluster 'someCluster' --service-name 'someService' --query 'taskArns[0]' --output 'text' \
+| xargs -oI '%%' \
+    aws ecs describe-tasks --cluster 'someCluster' --task '%%' \
+      --query 'tasks[].containers[?name==`someContainer`].image' --output 'text'
+
 # Delete services.
 aws ecs delete-service --cluster 'testCluster' --service 'testService' --force
 
@@ -148,7 +155,8 @@ while [[ $(aws ecs list-tasks --query 'taskArns' --output 'text' --cluster 'test
 # Restart tasks.
 # No real way to do that, just stop the tasks and new ones will be eventually started in their place.
 # To mimic a blue-green deployment, scale the service up by doubling its tasks, then down again to the normal amount.
-
+aws ecs update-service --cluster 'someCluster' --service 'someService' --desired-count '0' \
+&& aws ecs update-service --cluster 'someCluster' --service 'someService' --desired-count '1'
 ```
 
 </details>
@@ -1725,6 +1733,23 @@ Cost-saving measures:
   capacity provider.
 
     <details style='padding: 0 0 1rem 1rem'>
+      <summary> Percentage-like </summary>
+
+    ```json
+    {
+        "capacityProvider": "FARGATE",
+        "weight": 5
+    }
+    {
+        "capacityProvider": "FARGATE_SPOT",
+        "weight": 95
+    }
+    ```
+
+    </details>
+
+    <details style='padding: 0 0 1rem 1rem'>
+      <summary> Ratio-like </summary>
 
     ```json
     {
@@ -1775,6 +1800,13 @@ Specify a supported value for the task CPU and memory in your task definition.
 
 </details>
 
+## Cost-saving measures
+
+- Prefer using [spot capacity][effectively using spot instances in aws ecs for production workloads] for non-critical
+  services and tasks.
+- Consider applying for EC2 Instance and/or Compute Savings Plans if using EC2 capacity.<br/>
+  Consider applying for Compute Savings Plans if using Fargate capacity.
+
 ## Further readings
 
 - [Amazon Web Services]
@@ -1820,6 +1852,8 @@ Specify a supported value for the task CPU and memory in your task definition.
 - [Amazon ECS Service Discovery]
 - [AWS Fargate Pricing Explained]
 - [The Ultimate Beginner's Guide to AWS ECS]
+- [Amazon Amazon ECS launch types and capacity providers]
+- [Effectively Using Spot Instances in AWS ECS for Production Workloads]
 
 <!--
   Reference
@@ -1847,6 +1881,7 @@ Specify a supported value for the task CPU and memory in your task definition.
 [efs]: efs.md
 
 <!-- Upstream -->
+[Amazon Amazon ECS launch types and capacity providers]: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/capacity-launch-type-comparison.html
 [Amazon ECS capacity providers for the EC2 launch type]: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/asg-capacity-providers.html
 [Amazon ECS clusters for Fargate]: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/fargate-capacity-providers.html
 [Amazon ECS environment variables]: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/ecs-environment-variables.html
@@ -1865,6 +1900,7 @@ Specify a supported value for the task CPU and memory in your task definition.
 [AWS Fargate Spot Now Generally Available]: https://aws.amazon.com/blogs/aws/aws-fargate-spot-now-generally-available/
 [Centralized Container Logging with Fluent Bit]: https://aws.amazon.com/blogs/opensource/centralized-container-logging-fluent-bit/
 [ecs execute-command proposal]: https://github.com/aws/containers-roadmap/issues/1050
+[Effectively Using Spot Instances in AWS ECS for Production Workloads]: https://medium.com/@ankur.ecb/effectively-using-spot-instances-in-aws-ecs-for-production-workloads-d46985d0ae2d
 [EventBridge Scheduler]: https://docs.aws.amazon.com/scheduler/latest/UserGuide/what-is-scheduler.html
 [Example Amazon ECS task definition: Route logs to FireLens]: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/firelens-taskdef.html
 [fargate tasks sizes]: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/fargate-tasks-services.html#fargate-tasks-size

knowledge base/cloud computing/aws/eks.md

@@ -26,6 +26,7 @@
     1. [Identify common issues](#identify-common-issues)
     1. [The worker nodes fail to join the cluster](#the-worker-nodes-fail-to-join-the-cluster)
     1. [AWS ELB controller fails to get the region from the host's metadata](#aws-elb-controller-fails-to-get-the-region-from-the-hosts-metadata)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
     1. [Sources](#sources)
 
@@ -1413,6 +1414,13 @@ helm upgrade -i --repo 'https://aws.github.io/eks-charts' \
   --set 'vpcId'='vpc-01234567'
 ```
 
+## Cost-saving measures
+
+- Consider [using spot instances][building for cost optimization and resilience for eks with spot instances] for
+  non-critical workloads.
+- Consider applying for EC2 Instance and/or Compute Savings Plans if using EC2 worker nodes.
+  Consider applying for Compute Savings Plans if using Fargate profiles.
+
 ## Further readings
 
 - [Amazon Web Services]
@@ -1512,6 +1520,7 @@ helm upgrade -i --repo 'https://aws.github.io/eks-charts' \
 [aws eks create-nodegroup]: https://docs.aws.amazon.com/cli/latest/reference/eks/create-nodegroup.html
 [AWS Node Termination Handler]: https://github.com/aws/aws-node-termination-handler
 [awssupport-troubleshooteksworkernode runbook]: https://docs.aws.amazon.com/systems-manager-automation-runbooks/latest/userguide/automation-awssupport-troubleshooteksworkernode.html
+[Building for Cost optimization and Resilience for EKS with Spot Instances]: https://aws.amazon.com/blogs/compute/cost-optimization-and-resilience-eks-with-spot-instances/
 [choosing an amazon ec2 instance type]: https://docs.aws.amazon.com/eks/latest/userguide/choosing-instance-type.html
 [configure instance permissions required for systems manager]: https://docs.aws.amazon.com/systems-manager/latest/userguide/setup-instance-profile.html#instance-profile-policies-overview
 [create an amazon ebs csi driver iam role]: https://docs.aws.amazon.com/eks/latest/userguide/csi-iam-role.html

knowledge base/cloud computing/aws/opensearch.md

@@ -380,7 +380,7 @@ can manage.
 
 ## Cost-saving measures
 
-- Choose appropriate [instance types and sizes][supported instance types in amazon opensearch service].<br/>
+- Choose _appropriate_ [instance types and sizes][supported instance types in amazon opensearch service].<br/>
   Leverage the ability to select them to tailor the service offering to one's needs.
 
   > [OR1 instances][or1 storage for amazon opensearch service] **cannot** (currently?) be selected as master nodes.<br/>

knowledge base/cloud computing/aws/rds.md

@@ -28,6 +28,7 @@
     1. [ERROR: must be superuser to alter _X_ roles or change _X_ attribute](#error-must-be-superuser-to-alter-x-roles-or-change-x-attribute)
     1. [Transport fails asking for the remote user must have superuser, but it already does](#transport-fails-asking-for-the-remote-user-must-have-superuser-but-it-already-does)
     1. [The instance is unbearably slow](#the-instance-is-unbearably-slow)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
     1. [Sources](#sources)
 
@@ -103,6 +104,10 @@ Maintenance windows are paused when their DB instances are stopped.
 # Show details of RDS instances.
 aws rds describe-db-instances
 aws rds describe-db-instances --output 'json' --query "DBInstances[?(DBInstanceIdentifier=='master-prod')]"
+aws rds describe-db-instances --db-instance-identifier 'some-db-instance' \
+  --query 'DBInstances[0].InstanceCreateTime' --output 'text'
+aws rds describe-db-instances --db-instance-identifier 'some-db-instance' --output 'text' \
+  --query 'DBInstances[0]|join(``,[`postgresql://`,MasterUsername,`@`,Endpoint.Address,to_string(Endpoint.Port),`/`,DBname||`postgres`])'
 
 # Enable Performance Insights.
 aws rds modify-db-cluster --db-cluster-identifier 'staging-cluster' \
@@ -1073,6 +1078,16 @@ or write workloads and exceeds the instance type quotas.
 
 </details>
 
+## Cost-saving measures
+
+- Choose _appropriate_ instance types and sizes.
+- Prefer using [reserved instances][rds reserved instances] when one can stay on a single instance type for the whole
+  duration of the reservation.<br/>
+  Should the DB type **not** change in time, prefer _Standard RIs_. Otherwise, prefer _Convertible RIs_ for
+  flexibility.
+
+  RDS does **not** support Savings Plans at the time of writing.
+
 ## Further readings
 
 - [Working with DB instance read replicas]
@@ -1136,6 +1151,7 @@ or write workloads and exceeds the instance type quotas.
 [migrating databases using rds postgresql transportable databases]: https://aws.amazon.com/blogs/database/migrating-databases-using-rds-postgresql-transportable-databases/
 [Multi-AZ DB instance deployments for Amazon RDS]: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Concepts.MultiAZSingleStandby.html
 [pricing and data retention for performance insights]: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_PerfInsights.Overview.cost.html
+[RDS reserved instances]: https://aws.amazon.com/rds/reserved-instances/
 [Recommended alarms for RDS]: https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/Best_Practice_Recommended_Alarms_AWS_Services.html#RDS
 [renaming a db instance]: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_RenameInstance.html
 [Restoring a DB instance to a specified time for Amazon RDS]: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_PIT.html

knowledge base/cloud computing/aws/s3.md

@@ -3,6 +3,7 @@
 1. [TL;DR](#tldr)
 1. [Storage classes](#storage-classes)
 1. [Lifecycle configuration](#lifecycle-configuration)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
    1. [Sources](#sources)
 
@@ -215,6 +216,17 @@ actions. In such cases:
 
 Examples: [1][lifecycle  configuration examples], [2][s3 lifecycle rules examples]
 
+## Cost-saving measures
+
+- Prefer using lower storage classes for data that is not frequently accessed.<br/>
+  Lower storage classes have a minimum storage fee period **and** retrieval fees.
+- Consider using [Lifecycle configuration] to move down in storage tier that data that is not frequently accessed after
+  some time.
+- Prefer using [S3 Intelligent-Tiering][how s3 intelligent-tiering works] when not knowing how frequently data is
+  accessed.
+- Consider expiring old data after some time, if its retention is not needed.
+- Consider compressing data before uploading it.
+
 ## Further readings
 
 - [Amazon Web Services]
@@ -245,6 +257,8 @@ Examples: [1][lifecycle  configuration examples], [2][s3 lifecycle rules example
   -->
 
 <!-- In-article sections -->
+[Lifecycle configuration]: #lifecycle-configuration
+
 <!-- Knowledge base -->
 [amazon web services]: README.md
 [cli]: cli.md

knowledge base/cloud computing/aws/sagemaker.md

@@ -1,6 +1,7 @@
 # Sagemaker
 
 1. [TL;DR](#tldr)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
    1. [Sources](#sources)
 
@@ -11,6 +12,11 @@
 - Serverless Endpoints' backend use **a snapshot** of the Endpoint Configuration at the time each host is created.<br/>
   To make a serverless Endpoint use a new Configuration or Model, its hosts need to be replaced.
 
+## Cost-saving measures
+
+- Use a single endpoint for multiple models where it makes sense.
+- Delete endpoints when they are not used anymore.
+
 ## Further readings
 
 - [Amazon Web Services]

knowledge base/kubernetes/README.md

@@ -52,6 +52,7 @@ Hosted by the [Cloud Native Computing Foundation][cncf].
     1. [Run a command just before a Pod stops](#run-a-command-just-before-a-pod-stops)
 1. [Examples](#examples)
     1. [Create an admission webhook](#create-an-admission-webhook)
+1. [Cost-saving measures](#cost-saving-measures)
 1. [Further readings](#further-readings)
     1. [Sources](#sources)
 
@@ -1256,6 +1257,16 @@ you need:
 
 See the example's [README][create an admission webhook].
 
+## Cost-saving measures
+
+- Reconsider one's choices.<br/>
+  Does one really need a Kubernetes cluster? They introduce multiple redundancy, and have high complexity.<br/>
+  Consider the resources and maintenance efforts that will inevitably go into that.
+- Consider leveraging autoscaling.<br/>
+  See [Horizontal Pod Autoscaling] and [KEDA] to scale Pods depending on metrics.<br/>
+  See [Node Autoscaling][node scaling] to scale Nodes depending on number of Pods, node features, or resource
+  consumption.
+
 ## Further readings
 
 Usage:
@@ -1350,6 +1361,7 @@ Others:
 
 <!-- In-article sections -->
 [horizontal pod autoscaler]: #horizontal-pod-autoscaler
+[node scaling]: #node-scaling
 [vertical pod autoscaler]: #vertical-pod-autoscaler
 [pods]: #pods
 [privileged container vs privilege escalation]: #privileged-container-vs-privilege-escalation