VMworld 2013
Kevin Denton, Gilead Sciences
Jim Medeiros, VMware
Monica Sharma, VMware
Learn more about VMworld and register at http://www.vmworld.com/index.jspa?src=socmed-vmworld-slideshare
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...
Â
VMworld 2013: Tips and Tricks for Capacity Risk Assessment, Rightsizing and Planning
1. Tips and Tricks for Capacity Risk Assessment,
Rightsizing and Planning
Kevin Denton, Gilead Sciences
Jim Medeiros, VMware
Monica Sharma, VMware
VCM4992
#VCM4992
3. 3
Gilead - Overview
Gilead Sciences
⢠Growing, innovative leader in Research
based Biopharmaceutical
⢠Focus areas - HIV/AIDS, Hepatitis, Cancer,
Respiratory & Cardiovascular conditions
Goals ⢠Robust capacity planning based on
tangible data
⢠Forecast growth to know what capacity
is needed
4. 4
Gileadâs Challenges & Needs
Criteria for an Operations
Management Solution
ď§ No adequate capacity planning (Yearly fire drill)
ď§ No understanding of current utilization
ď§ No way to do adequate forecasting
Challenges
ď§ Drop & play â easy setup & management
ď§ Provides capabilities of showing Utilization,
Capacity management, Change management
and Forecasting
6. 6
Agenda: Tips and Tricks for vSphere Capacity Planning
vC Ops â Overview
Today & Roadmap Get Right Metrics Tune Policies Pick your Visuals
7. 7
Capacity Planning in vCenter Operations â Today
ď§ Ensure performance SLAs
ď§ Increase Utilization & Realize Savings
ď§ Plan better by what-if modeling
ď§ Policy driven Capacity views/dashboards
ď§ Optimization & Rightsizing
recommendations
ď§ Modeling of how many VMs can
fit & do I have enough
Do I have any capacity risk? Benefit
Benefit
Description
Can I improve utilization?
Do I have enough?
8. 8
Capacity Planning in vCenter Operations â Today
ď§ Ensure performance SLAs
ď§ Increase Utilization & Realize Savings
ď§ Plan better by what-if modeling
ď§ Policy driven Capacity views/dashboards
ď§ Optimization & Rightsizing
recommendations
ď§ Modeling of how many VMs can
fit & do I have enough
Do I have any capacity risk? Benefit
Benefit
Description
Can I improve utilization?
Do I have enough?
10. 10
Get the Right Metrics
16 GB- Total Allocated Capacity
2GB -What VM did not get (Contention)
8GB - What the VM got(Usage)
SQL VM
10GB- What the VM wants(Demand)
ď§ Demand is What the VM wants: Physical
resources an object might consume
w/o constraints
ď§ Demand = Usage (what VM gets)
+
Contention (What VM does not get)
⥠Check Time Resolution - Donât use one time
peak for planning, use rolled up avg over time
⢠Use BOTH: Allocation & Demand Models
⢠Use Allocation model to create a safe top line
E.g. fill VMs till cluster is at 200% ,then add
new host
⢠Use Demand model in conjunction to catch
unexpected bursts/peaks and prevent waste
⣠Compare actual demand vs. allocation
⢠To assess performance risk
⢠To show optimization potential & savings
ď§ Allocation - Amount of a resource that the
user configures
â Use Demand for capacity & performance
if Demand > Entitlement
⢠May have performance issues
⢠May be undersized (âStressedâ)
⢠Use Demand vs Consumed for Memory
Buffer The most a VM can get (Entitlement)
11. 11
Translate Your Operational âKnobsâ to vC Ops Policies
How would you like to
Manage Capacity Risk?
What are your goals to
Optimize your environment
Performance Higher utilization
Ignore Waste Higher densitysafe
PRODUCTION TEST-DEV
Configure Out-of-Box Policies
Production/Test Dev/UAT/IT-Apps etc
12. 12
Pick Your Visuals
Out of box
Custom
vSphere Dashboard Planning Views Canned Reports
Custom Templates Custom Heatmaps Custom Dashboards
13. 13
Resources available for you
1. VMworld slides
from
VMworld site
2. Custom Dashboards
from
VMware Management
Blog-Tech Tips
15. 15
Agenda: Tips & Tricks to Analyze Demand, Utilization & Risk
VM Growth Infra Burn Rate Capacity Risk
Monitor & Analyze
16. 16
How many of you have been tasked to
Monitor Infrastructure Utilization & Risk?
Audience Poll Question
17. 17
What Has Been My VM Growth Trend?
vC Ops vSphere UIď Planning ď Vm Capacity View vC Ops Custom UI->VM Count & Trend âby Cluster
â Metrics:
Use Total/Powered on
VM count
⥠Visuals:
Forecast trend to view
Risk
⢠View Growth
by Cluster, LOB, Geo etc.
18. 18
What Has Been My Infrastructure Utilization Trend?
⥠Visuals:
Breakdown
by cluster to view
Actual Demand
by Clusters
â Metrics:
Use Usable Capacity
vs. Total Capacity for
Planning decisions
(includes Buffers)
19. 19
How Well Is My Infrastructure Utilized Today?
⢠Under-utilized
Clusters â
fill or consolidate
⥠Stressed Clusters
with high Count
of VMs
â Used,
Remaining?
Metrics: VM Count,
Usable Capacity
20. 20
Which Clusters are at Capacity Risk & Why?
â Which clusters are
at Capacity Risk?
⢠Compare
Actual Demand
to Allocation
⥠Why?
- Out of Capacity?
- Will run out soon?
- Under-Sized?
- VM: Host Ratio
21. 21
Assess Risk Based on Your Policy
â Identify & Apply out of box Policies
⢠By Environment to manage Risk
⢠Production Policy
⢠Test-Dev Policy
⢠By Workload type for Right-sizing
⢠Ignore objects
⢠Batch Workloads
⢠Interactive/Server Workloads
⢠Optimized for 15/30 min SLA
⥠Translate your Knobs to Policies
⢠Allocation and Demand model
⢠Over-commit ratios(CPU, Mem)
⢠Thresholds for capacity risk
⢠Buffers
⢠Business hours
22. 22
What Do These Settings Impact & When?
â Dashboard - Time Remaining
& Capacity Remaining
calculated daily
⥠Planning Views â
Capacity Risk Details
view updates in real-time
23. 23
Which Datastores Are at Capacity Risk & Why?
Datastores at capacity
risk âcolor coded
Which VMs
Causing most waste?
24. 24
Which Top N VMs Are at Capacity Risk & Why?
VMs out of Capacity? Undersized VMs?
VMs out of Guest FS?
VMs running out of
capacity soon?
25. 25
Agenda: Tips and Tricks for Right-Sizing
Monitor & Analyze
Right-Size VMs ConclusionImprove Utilization
vC Ops â OverviewGileadâs Advantage
27. 27
Tips for Right Sizing VMs
â More vCPUs actually
slows down a VM
⥠(CPU Usage | Co-stop)
Trend this metric when
Usage is low but
Demand is high
Table for 2 â Just a minute please
Table for 10 â 20 minutes
28. 28
How Do Right Sizing Analytics Work?
Time
%Demand
Stress % Threshold
Current Capacity
Moments of Stress Summed Up as %
of Stress Zone Area
If Stress > 1%, show in under-sized VM listArea based Stress Analysis
⢠VM is considered
undersized/stressed when:
⢠Amount of CPU demand
peaks above 70% is more
than 1% of any 1 hour
70%
Time
%Demand
Current Capacity
Waste % Threshold
Moments of Wasted Summed
Up as % of Waste Zone Area
If Waste > 99%, show in list
⢠VM is considered oversized when:
⢠Amount of CPU demand below
above 30% is more than 1% of the
entire range(30 days)
30. 30
Step 2: Profile Workload & Apply Policy
Server Workload Profile:
⢠E.g. Exchange, AD, Citrix
⢠9-5 Usage pattern
⢠Account for many micro-
bursts in an hour
5 Minute CPU
Demand Average
Interactive Workload Profile:
⢠E.g. Web Servers
⢠Constantly busy
â Apply âInteractive Policyâ
⥠(Optional)Tune Settings
⢠To catch peaks
⢠Enable âStressâ
⢠Use buffers for erratic peaks
⢠Set sliding window = 1 hour
vSphere UI ď Operations ď All Metrics
31. 31
Step 2: Profile Batch Workload Type & Apply Policy
5 Minute CPU
Demand Average
Batch Workload Profile:
⢠E.g. Month end, Backup,
⢠Busy only for small bursts, idle most of the time.
Peak higher than avg
⢠Ensure sized for when it needs resources (4 hr SLA)
â Apply âBatch Workload Policyâ
⥠(Optional) Tune Settings:
⢠Narrow down business period
⢠Set âsliding windowâ for
expected duration
⢠If VM is idle for 28 days, it will
NOT be considered over-sized
32. 32
Step 3: Report Wasteful VMs with Usage Trends
Top N Over-sized VMs
Top N by Memory
Top N by CPU Usage
Trend Memory Demand
Trend CPU Demand
36. 36
Decide on Your Optimization Phases
1
20-50%
â Phase 1: Reclaim Waste
⢠Idle VMs
⢠Powered Off VMs
2
20%
⥠Phase 2: Increase Utilization
⢠Consolidate Under utilized
clusters
⢠Right-size Over-sized VMs
3
15%
⢠Phase 3: Increase Over-Commit
or Density âsafelyâ
⢠Assess potential density w/o
performance risk
37. 37
Phase 1: Reclaim Unused Resources (Waste)
â View Wasteful VMs
breakdown (Dashboard)
⥠Identify list of Idle, Powered
Off VMs in Planning
Views/Reports
38. 38
Phase 2: Consolidate Clusters
â Identify Under Utilized
Clusters to Consolidate
⥠Run what-if scenario
Select VMs from Under
utilized Cluster
Model if they will fit
in target cluster
⢠How many
Small
Medium
Large VMs
can fit in target cluster?
39. 39
Phase 3: Increase Over-commit Safely
â (Dashboard) Identify
optimal consolidation ratios
(Based on âDemandâ)
⥠Increase Over-commit
⢠Use allocation model for Memory
Risk management
⢠Increase Memory over-commit
by 5-15% and observe
⢠Set this in the Policy Settings 3c
40. 40
Conclusion & Takeaways
vCenter Operations Manager
enables you to improve your existing process to
Analyze, Optimize & Model future capacity needs
Gileadâs Advantage with vCenter Operations Manager
ďźRealized value within 3 months in production with vCenter Operations
ďźIdentified reclamation opportunities to realize savings
ďźGot improved insights to plan purchases for future growth
ďźGained more visibility into workloads to maintain performance & availability
41. 41
Other VMware Activities Related to This Session
ď§ HOL:
HOL-SDC-1301
Applied Cloud Operations
HOL-SDC-1304
vSphere Performance Optimization
44. Tips and Tricks for Capacity Risk Assessment,
Rightsizing and Planning
Kevin Denton, Gilead Sciences
Jim Medeiros, VMware
Monica Sharma, VMware
VCM4992
#VCM4992