1
COST OPTIMIZATION IN 5 PERSPECTIVES
REDUCE YOU CLOUD SPEND WITH 35%+ STRUCTURALLY!
AWS PREMIER CONSULTING PARTNER
ABOUT OBLIVION CLOUD CONTROL
Premier AWS consulting Partner
5+ years experience in Cloud Cost Management
Experienced market leading Cloud Cost management tools:
AWS Native, CloudHealth, CloudCheckr and Rightscale.
Cloud Cost Optimization scans and Cloud Cost Management services.
A TYPICAL CLOUD JOURNEY
Cloud
Operating
Model
2
3
Cloud
Foundation
(Landing zone)Cloud Compliancy
Control Framework
Cloud Business Case
Cloud
Discovery & -
Strategy
1
Cloud
Center of
Excellence
4
Training
Bootcamp
Cloud Capacity
(Cost Control)
Cloud
Migration
Plan
5
6
Cloud
Migration
Factory
7
Cloud
Optimization
/
Innovation
Application Discovery
and Assessments
Well Architected
Review(s)
Cloud Cost
Optimization Scan
4
COST OPTIMIZATION IN 5 PERSPECTIVES
REDUCE YOU CLOUD SPEND WITH 35%+ STRUCTURALLY!
AWS PREMIER CONSULTING PARTNER
Are you in control of
your cloud spend?
THE CLOUD BUSINESS CASE PROMISE
VS
The business case you
approved
Common reality
On-
Premises
Lift &
Shift
Instance
Right-Sizing
Improved
Elasticity
Measure,
Monitor,
Improve
Optimized
EC2
Storage
Optimization
Serverless
Architecture
Managed
Services
True AWS
Optimized
On-
Premises
Lift &
Shift
Post Migration
Period
CLOUD COST CONTROL IS CHALLENGING: 33% WASTED
Among Amazon Web Services® (AWS) users only 47 percent use AWS Reserved Instances.
Source: RightScale 2019
Source: RightScale 2019
Source: DevOps.com 2019
Source: RightScale 2019
Source: ZDNET 2019
Source :ZDNET
Traditionally IT spend was managed
via spend boards, and processes.
Today code generates the cost
Make the application teams
responsible and provide them the
insights and tools to take responsibility
Drive Accountability
Allocate cost to services, applications, departments
and more help to understand your costs
Understand the Costs
Visibility of generated cost by your cloud
platform is a fundamental first step
Gain visibility
The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in
Cost Analyses, Cost Optimization and Cost Management
Measure cloud cost and bring in the
ability to take control
Measure & Control
Analyze
Manage
Optim
ize
Establish Benchmarks
Define realistic benchmarks that will
help groups, on every level, contribute
to specific and company wide goals
THE CLOUD COST MANAGEMENT CYCLE
Make the application teams
responsible and provide them the
insights and tools to take responsibility
Drive Accountability
Allocate cost to services, applications, departments
and more help to understand your costs
Understand the Costs
Visibility of generated cost by your cloud
platform is a fundamental first step
Gain visibility
The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in
Cost Analyses, Cost Optimization and Cost Management
Measure cloud cost and bring in the
ability to take control
Measure & Control
Analyze
Manage
Optim
ize
Establish Benchmarks
Define realistic benchmarks that will
help groups, on every level, contribute
to specific and company wide goals
THE CLOUD COST MANAGEMENT CYCLE
Visibility
FROM TRANSPARENCY, AWARENESS, ACCOUNTABILITY, OWNERSHIP
TO CAPACITY MANAGEMENT
Every cloud resource/asset in the cloud
is bound to a unique Account that is
generated by the Cloud Provider and
assigned to one customer.
Cloud Resources/Assets
Resources/assets are labeled with
Application and or project identifiers
within the Accounts.
Accounts
Each one of the accounts is internally
associated with a Cost Center.
Cost Centers
Recharges the Departments/BUs by
allocating the costs of the accounts to the
respective Cost Centers being able to
provide more granularity.
IT/Finance
Can associate the costs of cloud
resources/assets based on business
initiatives in addition to accounts
Department/BU
The Cost Owner is responsible for the
Cost Centers and has better visibility to
the cost purpose.
Cost Owner
ACCOUNT BASED AND/OR TAGGING BASED
DIFFERENT ROLES, DEMANDS DIFFERENT INFORMATION
Show BackChargeback
Objective
OptimizationAwarenessAccounting
Purpose
DashboardReport
Type
UserPOCSCCM
Target
ApplicationProjectServiceAccountTeamBusiness
Unit
Cost Center
Category/Filter
Make the application teams
responsible and provide them the
insights and tools to take responsibility
Drive Accountability
Allocate cost to services, applications, departments
and more help to understand your costs
Understand the Costs
Visibility of generated cost by your cloud
platform is a fundamental first step
Gain visibility
The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in
Cost Analyses, Cost Optimization and Cost Management
Measure cloud cost and bring in the
ability to take control
Measure & Control
Analyze
Manage
Optim
ize
Establish Benchmarks
Define realistic benchmarks that will
help groups, on every level, contribute
to specific and company wide goals
THE CLOUD COST MANAGEMENT CYCLE
Visibility
AWS Resources such as EC2 Instances, EBS Volumes, Redshift and RDS
Clusters are often overprovisioned. With right sizing we focus on bringing
your AWS Resources to an efficient level.
SERVICE RIGHT SIZING
01
Up to 70% cost savings are achievable on specific AWS Services including
EC2, RDS, Redshift by committing your capacity for 1 or 3 years. We define
the appropriate strategy for you
PURCHASE MODEL OPTIMIZATION
02
Not all cost behavior can be analyzed according to checks from our
framework. With a Service Pattern Analyses unexpected cost pattern
deviations from your workloads normal patterns will be easily identified
and further analyzed.
SERVICE PATTERN ANALYSES
05
Each of AWS’s 185+ different services has different pricing characteristics.
Oblivion Cloud Control will identify misuse of a service from cost
perspective.
SERVICE CHARACTERISTICS
03
AWS launches over 1500 innovations yearly. Oblivion Cloud Control will
advise you on more efficient services for similar use cases.
SERVICE ARCHITECTURE
04
OUR FIVE PERSPECTIVES ON COST OPTIMIZATION
Amazon Linux on demand EC2Instances.info Frankfurt 14/02/2020
M5.xlarge (4/16)
$ 2.014,80
T3.large (2/8)
$ 940,82
M5.large (2/8)
$ 1.007,40
-50%
-6,6%
$ 10.739,80
Per 10 instances
Yearly Saving
01 SERVICE RIGHT SIZING
Things to consider
Eliminate ‘zombie assets’
Don’t provision for peak capacity
Consider ARM / AMD / T series
Reevaluate Provisioned capacity
Amazon Linux on demand EC2Instances.info Frankfurt 14/02/2020
M5.xlarge (4/16)
$ 2.014,80
T3.large (2/8)
$ 940,82
M5.large (2/8)
$ 1.007,40
-50%
-6,6%
$ 10.739,80
Per 10 instances
Yearly Saving
01 SERVICE RIGHT SIZING
Ireland - Amazon Linux Simply Monthly Calculator 14/02/2020
1 Year reserved instances
$6.519,20
Per 10 instances
Yearly Saving
-34,7%
m5.xlarge (4/16) m5.xlarge (4/16)
02 PURCHASE MODEL OPTIMIZATION
Things to consider
Ireland - Amazon Linux Simply Monthly Calculator 14/02/2020
1 Year reserved instances
$6.519,20
Per 10 instances
Yearly Saving
-34,7%
m5.xlarge (4/16) m5.xlarge (4/16)
Reserved Instances (up to 75%)
Spot Instances (variable up to 90%)
Savings plans (25% - 57%)
Subscription type (hourly, monthly
workspaces)
Enterprise Discount Programs (7-15%)
Reseller (value added) propositions
02 PURCHASE MODEL OPTIMIZATION
-3,6%
Latest instance type generation
m4.xlarge (2/8) m5.xlarge (2/8)
-50%
Availability zone requirements
db.r4.xlarge (4/30.5)
Region optimization
-69%
Storage type optimization
500 GB
PIOPs 1500
-7%
m5.large (2/8)
Frankfurt
m5.large (2/8)
Ireland
500 GB GP SSDdb.r4.xlarge (4/30.5)
Yearly Saving
$ 708,00
Per 10 instances
Yearly Saving
$ 46.860,00
Per 10 instances
Yearly Saving
$ 1.401,60
Per 10 instances
Yearly Saving
$ 14.640,00
Per 10 Volumes
03 SERVICE CHARACTERISTICS
-3,6%
Latest instance type generation
m4.xlarge (2/8) m5.xlarge (2/8)
-50%
Availability zone requirements
db.r4.xlarge (4/30.5)
Region optimization
-69%
Storage type optimization
500 GB
PIOPs 1500
-7%
m5.large (2/8)
Frankfurt
m5.large (2/8)
Ireland
500 GB GP SSDdb.r4.xlarge (4/30.5)
Yearly Saving
$ 708,00
Per 10 instances
Yearly Saving
$ 46.860,00
Per 10 instances
Yearly Saving
$ 1.401,60
Per 10 instances
Yearly Saving
$ 14.640,00
Per 10 Volumes
Things to consider
Latest instance type family
Storage types / storage tier
Multi AZ deployment Evaluation
RDS Storage autoscaling
Region optimization
DynamoDBIndex optimization
Scalability
TTL and Cache optimization
03 SERVICE CHARACTERISTICS
OS optimizationContainer workload optimization
Some examples
Amazon Elastic Kubernetes
Service
$ 1,701.31
EC2 t3.large, 20gb EBS
EKS on Fargate
$ 1,283.80
0.5 vCPU 1GB
ECS on Fargate
$ 407.80
0.5 vCPU 1GB
Yearly Saving
$ 5.255,40
Per 10 instances
-38%
m5.large (2/8)
Amazon Linux
$ 840,96
US East
m5.large (2/8)
RedHat Linux
$ 1366,55
US East
Things to consider
VPC endpoints
Apply caching
Build in elastisity
Offloading logging data
Fargate for containers
Open Source database engine
Amazon Linux
Glacier i.s.o. S3 for archives
Routing change
Query optimization
Code optimization
FSx i.s.o.
ec2 hosted fileserver
04 SERVICE ARCHITECTURE
OS optimizationContainer workload optimization
Some examples
Amazon Elastic Kubernetes
Service
$ 1,701.31
EC2 t3.large, 20gb EBS
EKS on Fargate
$ 1,283.80
0.5 vCPU 1GB
ECS on Fargate
$ 407.80
0.5 vCPU 1GB
Yearly Saving
$ 5.255,40
Per 10 instances
-38%
m5.large (2/8)
Amazon Linux
$ 840,96
US East
m5.large (2/8)
RedHat Linux
$ 1366,55
US East
04 SERVICE ARCHITECTURE
Abnormal behavior
Unexpected peaks
Measuring & Trendlines
$ 300 storage vs $ 3.500 requests
Inefficient application code
Increasing fleet of instances,
however
website visitors count stable.
Account Hacked, Crypto miners in
the AMI
Example cases
05 SERVICE PATTERN ANALYSES
A REAL LIFE SITUATION OF COMBINING THE PERSPECTIVES
M5.xlarge (4/16)
$ 2014,80
Frankfurt
M5.large (2/8)
$ 1007,40
Frankfurt
T3.large (2/8)
$ 940,82
Frankfurt
Rightsize
Service
Characteristics Reserve
T3.large (2/8)
$ 883,00
Ireland
T3.large (2/8)
$ 851,00
Ireland
(1-year RI
Make the application teams
responsible and provide them the
insights and tools to take responsibility
Drive Accountability
Allocate cost to services, applications, departments
and more help to understand your costs
Understand the Costs
Visibility of generated cost by your cloud
platform is a fundamental first step
Gain visibility
The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in
Cost Analyses, Cost Optimization and Cost Management
Measure cloud cost and bring in the
ability to take control
Measure & Control
Analyze
Manage
Optim
ize
Establish Benchmarks
Define realistic benchmarks that will
help groups, on every level, contribute
to specific and company wide goals
AUTOMATE YOUR CONTROLS
GOVERNANCE: COST CONTROLLERS & RESOURCE OWNERS
Cost
Controller
App
Team
Resource
Owner DevOps
Team
Project
Team
CCoE
Finance
DEFINE AND AUTOMATE YOUR COST CONTROL FRAMEWORK
PREVENTIVE DETECTIVE CORRECTIVE
- Documentation
- Code Peer reviews
- Log everything
- Use of restrictive IAM policies
(ie. “deny” on creating
CloudTrail trails)
- …
- AWS Config
- In combination with use of
alarms/alerts/forecasts
- Dashboards and reports
- Schedule regular reviews of
cost and usage metrics
- Automated checks (e.g aging
snapshots, unattached EBS
volumes, old instance
generations, …)
- …
- (automated) resource
optimizations
- Training
- Cost aware culture
- …
28
WRAP-UP
AWS PREMIER CONSULTING PARTNER
COST OPTIMIZATION FRAMEWORK
EC2
ECS EKS
Lambda
Fargate
SNS
• R) Apply Right sizing
• R) Decommision Zombie instances
• P) Savings Plans / Reserved instances
• P) Consider Spot instances
• S) Latest instance Family generation
• S) Use of burstable instances (Tseries)
• S) Use of AMD or ARM based
instances
• A) Apply Scaling Policies
• A) Switch to open source OS Versions
optimization (amz lin)
• R) Optimize Memory
• P) Provisioned Concurrency for high
traffic
• A) Consider offload to step functions
• P) Savings Plans / Reserved Instances
• P) Consider Spot instances
• S) use Capacity Provider to scale
efficient
• A) Go Serverless with ECS Fargate
• P) Consider Savings Plans
• A) Consider ECS Fargate
• A) Go Serverless with EKS Fargate
• P) Consider Savings Plans
• P) Use Fargate spot for non prod
• R) Delete unrelevant buckets/objects
• S) Apply Lifecycle policies with
versioning
• S) Optimize Storage tier or consider
intelligent tiering
• S) Choose the right AWS region to
reduce inter region traffic or
consider CloudFront
• A) Consider Glacier for cold data
• B) analyse API cost for operation of
files
EBS
S3 FSx
for Windows
EFS
Storage
Gateway
Backup
• R) Right size provisioned EBS volumes
• R) Decommision unattached volumes
• R) Delete old snapshots
• S) Reevaluate prov. IOPS volumes for
general purpose EBS volumes
• A) Offload cold storage to cheaper
storage
• R) Right size provisioned capacity
• R) Right size toughput volume
• S) Single AZ deployment for Non
Prod
• A) Offload cold storage to cheaper
storage
• A) Apply dedup/compression
AURORA DYNAMO DBRDS
ELASTIC CACHEREDSHIFT
• Apply Right sizing
• ….
• Apply Right sizing
• Consider reservations
• R) Apply Right sizing
• P) Consider reservations
• S) Single AZ deployment for Non
Prod
• S) Use RDS storage autoscaling
• S) Fit for purpose storage type
• A) Switch to open source DB Engines
• A) Consider Aurora serverless for
MYSQL
• Apply Autoscaling iso Provisioned
• Capacity Reservation
• Usage vs provisioned
• Over-provisioning to handle hot
partitions
• Provisioned throughput
• Indexed data storage
• Use of caching
• Global Tables
• On-demand backup
• Continuous backups//point-in time
recovery
• Table restores
• Apply Right sizing
• ….
VPC
VPN Transit Gateway
CloudFront
Route 53
SNS
• R )Release Unattached Elastic IP’s
• A) Prevent inter AZ / region traffic
• A) VPC Endpoints for heavy traffic to
S3 and Dynamo DB
• A) Reevaluate NAT Gatway setup • S) Optimize Cache TTL and regions
• A) Go • P) C
• A) Consoliate resolver endpoints
• S) Restrict direct S3 access (Origin
Access)
NAT Gateway
• R)
API Gateway
• P)
• R) /
• S) g
Cloud Watch
Guard Duty Inspector
Cloud Trail
KMS
Config
• R) R… • R) …
Certificate
Manager
• R) /
• S) g
• R) Remove the un-necessary logs
• S)
• A) Offload to alternative logging
service
• R) /
• S) g
SQS
SNS
APPSYNC
STEP
FUNCTIONS
App Stream Workspaces
• R) Apply)
• R) Right size compute instance
• P) reevaluate Monthly vs hourly
payment & RI
• A) consider App stream for some
usecases
EMR Athena QuickSight
Glue Kinesis
Elastic
Search
EC2
• R) Apply Right sizing
• R) Decommision Zombie instances
• P) Savings Plans / Reserved instances
• P) Consider Spot instances
• S) Latest instance Family generation
• S) Use of burstable instances (Tseries)
• S) Use of AMD or ARM based instances
• A) Apply Scaling Policies
• A) Switch to open source OS Versions
optimization (amz lin)
RDS
• R) Apply Right sizing
• P) Consider reservations
• S) Single AZ deployment for Non Prod
• S) Use RDS storage autoscaling
• S) Fit for purpose storage type
• A) Switch to open source DB Engines
• A) Consider Aurora serverless for MYSQL
Stripped version
START YOUR CONVERSATION
Request your set via:
jeroen.jacobs@oblcc.com
+31 6 29 52 60 92
TAKE AWAYS
Start with cost insights and awareness (Tagging is Key)
Focus on the quick wins (right size, zombie asset reduction, reservations)
Approach Cloud Cost management from a TCO perspective
Drive the full cloud cost management circle
Request help, there is always a business case for cost optimizations
32
END OF PRESENTATION
Request your set via:
jeroen.jacobs@oblcc.com
+31 6 29 52 60 92

AWS Cost Optimization in 5 Perspective

  • 1.
    1 COST OPTIMIZATION IN5 PERSPECTIVES REDUCE YOU CLOUD SPEND WITH 35%+ STRUCTURALLY! AWS PREMIER CONSULTING PARTNER
  • 2.
    ABOUT OBLIVION CLOUDCONTROL Premier AWS consulting Partner 5+ years experience in Cloud Cost Management Experienced market leading Cloud Cost management tools: AWS Native, CloudHealth, CloudCheckr and Rightscale. Cloud Cost Optimization scans and Cloud Cost Management services.
  • 3.
    A TYPICAL CLOUDJOURNEY Cloud Operating Model 2 3 Cloud Foundation (Landing zone)Cloud Compliancy Control Framework Cloud Business Case Cloud Discovery & - Strategy 1 Cloud Center of Excellence 4 Training Bootcamp Cloud Capacity (Cost Control) Cloud Migration Plan 5 6 Cloud Migration Factory 7 Cloud Optimization / Innovation Application Discovery and Assessments Well Architected Review(s) Cloud Cost Optimization Scan
  • 4.
    4 COST OPTIMIZATION IN5 PERSPECTIVES REDUCE YOU CLOUD SPEND WITH 35%+ STRUCTURALLY! AWS PREMIER CONSULTING PARTNER
  • 5.
    Are you incontrol of your cloud spend?
  • 6.
    THE CLOUD BUSINESSCASE PROMISE VS The business case you approved Common reality On- Premises Lift & Shift Instance Right-Sizing Improved Elasticity Measure, Monitor, Improve Optimized EC2 Storage Optimization Serverless Architecture Managed Services True AWS Optimized On- Premises Lift & Shift Post Migration Period
  • 7.
    CLOUD COST CONTROLIS CHALLENGING: 33% WASTED Among Amazon Web Services® (AWS) users only 47 percent use AWS Reserved Instances. Source: RightScale 2019 Source: RightScale 2019 Source: DevOps.com 2019 Source: RightScale 2019 Source: ZDNET 2019 Source :ZDNET
  • 8.
    Traditionally IT spendwas managed via spend boards, and processes. Today code generates the cost
  • 9.
    Make the applicationteams responsible and provide them the insights and tools to take responsibility Drive Accountability Allocate cost to services, applications, departments and more help to understand your costs Understand the Costs Visibility of generated cost by your cloud platform is a fundamental first step Gain visibility The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in Cost Analyses, Cost Optimization and Cost Management Measure cloud cost and bring in the ability to take control Measure & Control Analyze Manage Optim ize Establish Benchmarks Define realistic benchmarks that will help groups, on every level, contribute to specific and company wide goals THE CLOUD COST MANAGEMENT CYCLE
  • 10.
    Make the applicationteams responsible and provide them the insights and tools to take responsibility Drive Accountability Allocate cost to services, applications, departments and more help to understand your costs Understand the Costs Visibility of generated cost by your cloud platform is a fundamental first step Gain visibility The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in Cost Analyses, Cost Optimization and Cost Management Measure cloud cost and bring in the ability to take control Measure & Control Analyze Manage Optim ize Establish Benchmarks Define realistic benchmarks that will help groups, on every level, contribute to specific and company wide goals THE CLOUD COST MANAGEMENT CYCLE Visibility
  • 11.
    FROM TRANSPARENCY, AWARENESS,ACCOUNTABILITY, OWNERSHIP TO CAPACITY MANAGEMENT Every cloud resource/asset in the cloud is bound to a unique Account that is generated by the Cloud Provider and assigned to one customer. Cloud Resources/Assets Resources/assets are labeled with Application and or project identifiers within the Accounts. Accounts Each one of the accounts is internally associated with a Cost Center. Cost Centers Recharges the Departments/BUs by allocating the costs of the accounts to the respective Cost Centers being able to provide more granularity. IT/Finance Can associate the costs of cloud resources/assets based on business initiatives in addition to accounts Department/BU The Cost Owner is responsible for the Cost Centers and has better visibility to the cost purpose. Cost Owner ACCOUNT BASED AND/OR TAGGING BASED
  • 12.
    DIFFERENT ROLES, DEMANDSDIFFERENT INFORMATION Show BackChargeback Objective OptimizationAwarenessAccounting Purpose DashboardReport Type UserPOCSCCM Target ApplicationProjectServiceAccountTeamBusiness Unit Cost Center Category/Filter
  • 13.
    Make the applicationteams responsible and provide them the insights and tools to take responsibility Drive Accountability Allocate cost to services, applications, departments and more help to understand your costs Understand the Costs Visibility of generated cost by your cloud platform is a fundamental first step Gain visibility The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in Cost Analyses, Cost Optimization and Cost Management Measure cloud cost and bring in the ability to take control Measure & Control Analyze Manage Optim ize Establish Benchmarks Define realistic benchmarks that will help groups, on every level, contribute to specific and company wide goals THE CLOUD COST MANAGEMENT CYCLE Visibility
  • 14.
    AWS Resources suchas EC2 Instances, EBS Volumes, Redshift and RDS Clusters are often overprovisioned. With right sizing we focus on bringing your AWS Resources to an efficient level. SERVICE RIGHT SIZING 01 Up to 70% cost savings are achievable on specific AWS Services including EC2, RDS, Redshift by committing your capacity for 1 or 3 years. We define the appropriate strategy for you PURCHASE MODEL OPTIMIZATION 02 Not all cost behavior can be analyzed according to checks from our framework. With a Service Pattern Analyses unexpected cost pattern deviations from your workloads normal patterns will be easily identified and further analyzed. SERVICE PATTERN ANALYSES 05 Each of AWS’s 185+ different services has different pricing characteristics. Oblivion Cloud Control will identify misuse of a service from cost perspective. SERVICE CHARACTERISTICS 03 AWS launches over 1500 innovations yearly. Oblivion Cloud Control will advise you on more efficient services for similar use cases. SERVICE ARCHITECTURE 04 OUR FIVE PERSPECTIVES ON COST OPTIMIZATION
  • 15.
    Amazon Linux ondemand EC2Instances.info Frankfurt 14/02/2020 M5.xlarge (4/16) $ 2.014,80 T3.large (2/8) $ 940,82 M5.large (2/8) $ 1.007,40 -50% -6,6% $ 10.739,80 Per 10 instances Yearly Saving 01 SERVICE RIGHT SIZING
  • 16.
    Things to consider Eliminate‘zombie assets’ Don’t provision for peak capacity Consider ARM / AMD / T series Reevaluate Provisioned capacity Amazon Linux on demand EC2Instances.info Frankfurt 14/02/2020 M5.xlarge (4/16) $ 2.014,80 T3.large (2/8) $ 940,82 M5.large (2/8) $ 1.007,40 -50% -6,6% $ 10.739,80 Per 10 instances Yearly Saving 01 SERVICE RIGHT SIZING
  • 17.
    Ireland - AmazonLinux Simply Monthly Calculator 14/02/2020 1 Year reserved instances $6.519,20 Per 10 instances Yearly Saving -34,7% m5.xlarge (4/16) m5.xlarge (4/16) 02 PURCHASE MODEL OPTIMIZATION
  • 18.
    Things to consider Ireland- Amazon Linux Simply Monthly Calculator 14/02/2020 1 Year reserved instances $6.519,20 Per 10 instances Yearly Saving -34,7% m5.xlarge (4/16) m5.xlarge (4/16) Reserved Instances (up to 75%) Spot Instances (variable up to 90%) Savings plans (25% - 57%) Subscription type (hourly, monthly workspaces) Enterprise Discount Programs (7-15%) Reseller (value added) propositions 02 PURCHASE MODEL OPTIMIZATION
  • 19.
    -3,6% Latest instance typegeneration m4.xlarge (2/8) m5.xlarge (2/8) -50% Availability zone requirements db.r4.xlarge (4/30.5) Region optimization -69% Storage type optimization 500 GB PIOPs 1500 -7% m5.large (2/8) Frankfurt m5.large (2/8) Ireland 500 GB GP SSDdb.r4.xlarge (4/30.5) Yearly Saving $ 708,00 Per 10 instances Yearly Saving $ 46.860,00 Per 10 instances Yearly Saving $ 1.401,60 Per 10 instances Yearly Saving $ 14.640,00 Per 10 Volumes 03 SERVICE CHARACTERISTICS
  • 20.
    -3,6% Latest instance typegeneration m4.xlarge (2/8) m5.xlarge (2/8) -50% Availability zone requirements db.r4.xlarge (4/30.5) Region optimization -69% Storage type optimization 500 GB PIOPs 1500 -7% m5.large (2/8) Frankfurt m5.large (2/8) Ireland 500 GB GP SSDdb.r4.xlarge (4/30.5) Yearly Saving $ 708,00 Per 10 instances Yearly Saving $ 46.860,00 Per 10 instances Yearly Saving $ 1.401,60 Per 10 instances Yearly Saving $ 14.640,00 Per 10 Volumes Things to consider Latest instance type family Storage types / storage tier Multi AZ deployment Evaluation RDS Storage autoscaling Region optimization DynamoDBIndex optimization Scalability TTL and Cache optimization 03 SERVICE CHARACTERISTICS
  • 21.
    OS optimizationContainer workloadoptimization Some examples Amazon Elastic Kubernetes Service $ 1,701.31 EC2 t3.large, 20gb EBS EKS on Fargate $ 1,283.80 0.5 vCPU 1GB ECS on Fargate $ 407.80 0.5 vCPU 1GB Yearly Saving $ 5.255,40 Per 10 instances -38% m5.large (2/8) Amazon Linux $ 840,96 US East m5.large (2/8) RedHat Linux $ 1366,55 US East Things to consider VPC endpoints Apply caching Build in elastisity Offloading logging data Fargate for containers Open Source database engine Amazon Linux Glacier i.s.o. S3 for archives Routing change Query optimization Code optimization FSx i.s.o. ec2 hosted fileserver 04 SERVICE ARCHITECTURE
  • 22.
    OS optimizationContainer workloadoptimization Some examples Amazon Elastic Kubernetes Service $ 1,701.31 EC2 t3.large, 20gb EBS EKS on Fargate $ 1,283.80 0.5 vCPU 1GB ECS on Fargate $ 407.80 0.5 vCPU 1GB Yearly Saving $ 5.255,40 Per 10 instances -38% m5.large (2/8) Amazon Linux $ 840,96 US East m5.large (2/8) RedHat Linux $ 1366,55 US East 04 SERVICE ARCHITECTURE
  • 23.
    Abnormal behavior Unexpected peaks Measuring& Trendlines $ 300 storage vs $ 3.500 requests Inefficient application code Increasing fleet of instances, however website visitors count stable. Account Hacked, Crypto miners in the AMI Example cases 05 SERVICE PATTERN ANALYSES
  • 24.
    A REAL LIFESITUATION OF COMBINING THE PERSPECTIVES M5.xlarge (4/16) $ 2014,80 Frankfurt M5.large (2/8) $ 1007,40 Frankfurt T3.large (2/8) $ 940,82 Frankfurt Rightsize Service Characteristics Reserve T3.large (2/8) $ 883,00 Ireland T3.large (2/8) $ 851,00 Ireland (1-year RI
  • 25.
    Make the applicationteams responsible and provide them the insights and tools to take responsibility Drive Accountability Allocate cost to services, applications, departments and more help to understand your costs Understand the Costs Visibility of generated cost by your cloud platform is a fundamental first step Gain visibility The Oblivion Cloud Control Cost Cost Management Methodology approaches in five steps grouped in Cost Analyses, Cost Optimization and Cost Management Measure cloud cost and bring in the ability to take control Measure & Control Analyze Manage Optim ize Establish Benchmarks Define realistic benchmarks that will help groups, on every level, contribute to specific and company wide goals AUTOMATE YOUR CONTROLS
  • 26.
    GOVERNANCE: COST CONTROLLERS& RESOURCE OWNERS Cost Controller App Team Resource Owner DevOps Team Project Team CCoE Finance
  • 27.
    DEFINE AND AUTOMATEYOUR COST CONTROL FRAMEWORK PREVENTIVE DETECTIVE CORRECTIVE - Documentation - Code Peer reviews - Log everything - Use of restrictive IAM policies (ie. “deny” on creating CloudTrail trails) - … - AWS Config - In combination with use of alarms/alerts/forecasts - Dashboards and reports - Schedule regular reviews of cost and usage metrics - Automated checks (e.g aging snapshots, unattached EBS volumes, old instance generations, …) - … - (automated) resource optimizations - Training - Cost aware culture - …
  • 28.
  • 29.
    COST OPTIMIZATION FRAMEWORK EC2 ECSEKS Lambda Fargate SNS • R) Apply Right sizing • R) Decommision Zombie instances • P) Savings Plans / Reserved instances • P) Consider Spot instances • S) Latest instance Family generation • S) Use of burstable instances (Tseries) • S) Use of AMD or ARM based instances • A) Apply Scaling Policies • A) Switch to open source OS Versions optimization (amz lin) • R) Optimize Memory • P) Provisioned Concurrency for high traffic • A) Consider offload to step functions • P) Savings Plans / Reserved Instances • P) Consider Spot instances • S) use Capacity Provider to scale efficient • A) Go Serverless with ECS Fargate • P) Consider Savings Plans • A) Consider ECS Fargate • A) Go Serverless with EKS Fargate • P) Consider Savings Plans • P) Use Fargate spot for non prod • R) Delete unrelevant buckets/objects • S) Apply Lifecycle policies with versioning • S) Optimize Storage tier or consider intelligent tiering • S) Choose the right AWS region to reduce inter region traffic or consider CloudFront • A) Consider Glacier for cold data • B) analyse API cost for operation of files EBS S3 FSx for Windows EFS Storage Gateway Backup • R) Right size provisioned EBS volumes • R) Decommision unattached volumes • R) Delete old snapshots • S) Reevaluate prov. IOPS volumes for general purpose EBS volumes • A) Offload cold storage to cheaper storage • R) Right size provisioned capacity • R) Right size toughput volume • S) Single AZ deployment for Non Prod • A) Offload cold storage to cheaper storage • A) Apply dedup/compression AURORA DYNAMO DBRDS ELASTIC CACHEREDSHIFT • Apply Right sizing • …. • Apply Right sizing • Consider reservations • R) Apply Right sizing • P) Consider reservations • S) Single AZ deployment for Non Prod • S) Use RDS storage autoscaling • S) Fit for purpose storage type • A) Switch to open source DB Engines • A) Consider Aurora serverless for MYSQL • Apply Autoscaling iso Provisioned • Capacity Reservation • Usage vs provisioned • Over-provisioning to handle hot partitions • Provisioned throughput • Indexed data storage • Use of caching • Global Tables • On-demand backup • Continuous backups//point-in time recovery • Table restores • Apply Right sizing • …. VPC VPN Transit Gateway CloudFront Route 53 SNS • R )Release Unattached Elastic IP’s • A) Prevent inter AZ / region traffic • A) VPC Endpoints for heavy traffic to S3 and Dynamo DB • A) Reevaluate NAT Gatway setup • S) Optimize Cache TTL and regions • A) Go • P) C • A) Consoliate resolver endpoints • S) Restrict direct S3 access (Origin Access) NAT Gateway • R) API Gateway • P) • R) / • S) g Cloud Watch Guard Duty Inspector Cloud Trail KMS Config • R) R… • R) … Certificate Manager • R) / • S) g • R) Remove the un-necessary logs • S) • A) Offload to alternative logging service • R) / • S) g SQS SNS APPSYNC STEP FUNCTIONS App Stream Workspaces • R) Apply) • R) Right size compute instance • P) reevaluate Monthly vs hourly payment & RI • A) consider App stream for some usecases EMR Athena QuickSight Glue Kinesis Elastic Search EC2 • R) Apply Right sizing • R) Decommision Zombie instances • P) Savings Plans / Reserved instances • P) Consider Spot instances • S) Latest instance Family generation • S) Use of burstable instances (Tseries) • S) Use of AMD or ARM based instances • A) Apply Scaling Policies • A) Switch to open source OS Versions optimization (amz lin) RDS • R) Apply Right sizing • P) Consider reservations • S) Single AZ deployment for Non Prod • S) Use RDS storage autoscaling • S) Fit for purpose storage type • A) Switch to open source DB Engines • A) Consider Aurora serverless for MYSQL Stripped version
  • 30.
    START YOUR CONVERSATION Requestyour set via: jeroen.jacobs@oblcc.com +31 6 29 52 60 92
  • 31.
    TAKE AWAYS Start withcost insights and awareness (Tagging is Key) Focus on the quick wins (right size, zombie asset reduction, reservations) Approach Cloud Cost management from a TCO perspective Drive the full cloud cost management circle Request help, there is always a business case for cost optimizations
  • 32.
    32 END OF PRESENTATION Requestyour set via: jeroen.jacobs@oblcc.com +31 6 29 52 60 92