Anecdotal numbers suggest that more than 40% compute resources are under utilized -- from unused cloud instances to virtual machines running on bare-metal. Hundreds of QA & Dev nodes to thousands of production instances could be shutdown, and brought back to the same state on demand. That's what cloud is about -- agility and efficiency, but our on-premise datacenter habits have migrated to the cloud as well.
Calm's DevOps automation platform helps fix our old habits. Calm provides a single pane of glass across cloud and on-premise, integrating with Chef, Puppet and Docker ecosystems. The single pane of glass enables orchestration, cost-control and on-demand provisioning.
Cost Control Across Cloud, On-Premise and VM Computers by Mark Lavi, Calm.io
1. Cost Control Across Containers,
Cloud, On-Premise, and VMs
Mark Lavi
Technology Evangelist
mark@calm.io
2. Problems:
• Explosive
Growth
• Cost Control
Challenges
Agenda
Strategies:
• Controlling
Costs
• How Much
Can You
Save?
Solutions:
• Using Calm to
Save
• Calm Demo
@booth G10
5. • Lack of Visibility
• Transferring Real-world Budgets and Controls
• Application Architecture Issues
Cost Control Challenges
Why do we have monthly sticker shock?
6. • Teams provision resources by hand and never turn them off
or release associated resources
• Resource ownership unknown:
• No meta data or naming convention enforcement
• Tracking and allocation of resources using spreadsheets
• Resource overview requires credentials to each account in
every provider
Lack of Visibility
Without oversight, we react to problems.
7. It worked before, why not now?
• Tool fragmentation prevents global project, team, or department
controls; people consume as much as they want
• Management would like to predict and enforce cloud spend so that
budgets can be planned by project, team, or department
• Multiple clouds and hybrid infrastructure prevent consistent charge
back model
Transferring Real-world Budgets
8. Can’t we have universal governance?
How do we enable global notification and approval policies for any:
• Business property, e.g.:
team, budget, provider account, etc.
• Application life cycle event, e.g.:
deployment, maintenance, upgrade, restart, etc.
Transferring Real-world Controls
9. Carrying forward traditional approaches = technical debt!
Traditional resource planning sizes for maximum capacity:
• Some to most capacity wasted every day
• Static populations adjusted manually, reactively
• You pay for your consumption and waste!
Traditional architectures represent friction on engineering delivery:
• Time to market for new features + problem fixes is critical
• Monolithic resources remain underutilized and require change
control windows, human error, and downtime
• Slow delivery causes customers to go elsewhere!
Application Architecture Issues
11. • Governance and Compliance
• Team Budgets
• Application Expiration
• Elastic Scale and Containerization
Controlling Costs
From lack of visibility & controls to reinforcing dynamic policies.
12. We want to manage our people and operations.
1. Role Based Access Control (RBAC):
• Administrators versus Designers versus Consumers
2. Audit Trails:
• Who did what, where, and when?
• How long did it take, how much did it cost?
Expected Results:
• Self-service deployments, democratizing operations
• Delegatable, point in time forensics and troubleshooting
• Streaming Logs for real-time analysis, predictive resource analytics
• Integration of all people, tools, and platforms
Governance and Compliance
13. A team is as strong as its weakest link.
1. Teams + user roles can govern deployments and life cycle operations
2. Team budgets can control consumption
3. Dashboard overview of the current team spend
Expected Results:
• Budgets = quotas: no more unchecked team deployments
• Visibility of team and application consumption
• Savings: a fixed amount constrains spending
Team Budgets
14. Dynamic policy enforcement for apps!
1. Configure application deployments to expire
2. Restrict application deployments to Developer & QA teams
3. Configure policies for team notification and approval
Expected Results:
• Self-service deployments and notifications, democratizing operations
• Expiration controls deployment lifecycle, sooner than team budgets!
• Evolution to ephemeral infrastructure for ephemeral deployments
• Savings: variable compared to perpetual deployments
Application Expiration
15. Dynamic policy enforcement for resources!
1. Infrastructure population should be controlled via monitoring to trigger
application tier provisioning operations
2. Application density should increase to resource capacity
Expected Results:
• Resource capacity becomes dynamic, elastic based on demand
• Containers enable new architectural practices:
• Immutable infrastructure and micro-services
• Transition from monolithic hosts to multiple applications per host
• Savings: variable with demand and co-location
Elastic Scale & Containerization
17. Fixed Savings:
• Governance and Compliance
• Team Budgets
Variable % Savings:
• Application Expiration
• Elastic Scale
• Containerization
How Much Money Can You Save?
Why do we not have monthly savings sticker shock?
18. Policy communication is priceless.
Notifications across projects, teams, and providers can give visibility to spend
and lifecycle events for teams or management. Examples:
• Notify when overall spend exceeds US$30K on AWS
• Require operations approval every time someone resets a VM in
production
• Require VP Engineering approval for a hotfix update to production
• Notify sales, marketing, support departments when an upgrade
completes
Savings depends on controlled versus uncontrolled cost delta per period
Savings: Global Governance
19. Let’s measure this period and plan the next.
Budgets can help the team and management discuss remaining and adjust next
period resource quotas across hybrid IT
Set and forget per period example budget:
• Trigger a notification at 85% spend
• Team cannot exceed US$10K deployment costs
Savings depends on controlled versus uncontrolled cost delta per period
Savings: Team Budgets
20. Stretch your saving!
Using a typical sinusoidal demand model for each local region:
• weekday vs. weekend cycle: 30% off is typical
• business hour vs off-business hours on weekdays: 50% typical
We can achieve 40+% regional savings (resolution dependent)!
( (50% * 5 days/week * 4 weeks * 2/3 day)
+(70% * 2 days/week * 4 weeks)
) / (30 days/month) = (6.66 + 5.6) / 30
= 40.86% monthly savings
Savings: Elastic Scale
21. Contain your saving!
Assuming a multi-tiered application architecture, e.g.:
1. web tier
2. business logic tier
We can co-locate both application tiers on each container host. By placing both
tiers per container host, we can achieve 50% savings compared to monolithic
tiers
Application dependent savings can increase using container workload
schedulers!
Savings: Containerization
22. Automate the end of the day.
Expiration can control unchecked deployments, encouraging evolution to
ephemeral infrastructure
Simply re-deploying each day and expiring 8 business hours later would equal
66% daily savings compared to perpetual deployments for non-production work
When we count weekends off, savings increase further!
Savings: Application Expiration
24. Show me the money!
Fixed savings compared to uncontrolled (can vary):
• Governance and Compliance
• Team Budgets
Variable % savings per team, period, and work load:
• 66%+ = Application Expiration
• 50% = Containerization
• 40%+ = Elastic Scale
Usage is exploding: save money before it is spent!
Using Calm to Save
25. Ready for the demo? Please visit Booth G10!
Calm is a DevOps automation platform that bring together people, tools, and
providers via visual infrastructure as code that orchestrates life cycle operations
and policies.
Calm Demo: Role Based Access Control
of a hybrid app deployment (containers and VMs)
under a team budget with expiration
More demos at Booth G10 including Docker Swarm + K8s!
Visit Calm.io to Save
26. Calm: Cost Controls
• Budgets: for team/project/etc.
• Visibility: for deployments
• Policy: lifecycle event approval
28. Calm: Hybrid Deployment
• Visual infrastructure-as-code
• Model your app architecture and
orchestrate dependencies
• Span across containers, VMs,
bare metal on public & private
clouds and data centers
• Refactor and migrate traditional
workloads to containers!
29. Calm: Lifecycle Management
• Run book automation
• Orchestrate complex ops
• Rolling/Canary updates
• Flexible autoscale
31. Slide subtitle if needed
The quick brown fox jumps over the lazy
dog. The quick brown fox jumps over the
lazy dog. The quick brown fox jumps over
the lazy dog.
• Bulleted list one
• List item two
• And item number three
Body Slide
32. Slide subtitle if needed
The quick brown fox jumps over the lazy
dog. The quick brown fox jumps over the
lazy dog. The quick brown fox jumps over
the lazy dog.
• Bulleted list one
• List item two
• And item number three
Alt Background Body Slide