VMworld 2012 Live Blog of OPS-CIM1926 Must Know Design Considerations for Planning Capacity When You are 50 Percent or More Virtualized

This is the Live Blog of OPS-CIM1926 − 5 Must Know Design Considerations for Planning Capacity When you are 50 Percent or More Virtualized. You’ll find my recap of the session below.

Speakers:

  • Samuel McBride – VMware, Inc.
  • Monica Sharma, VMware, Inc.

This session will discuss capacity planning. Monica Sharma is speaking now. The right management tools allow the ratio of VM per Admin to be much more efficient. We can manage more with fewer resources.

Capacity problem areas can get amplified with higher consolidation.

  • Performance Triage / Reactive Capacity Remediation – Over-commitment / Reservations tend to be common in larger environments
  • It is difficult to cross-reference multiple tools to obtain the most relevant information needed to troubleshoot
  • This leads to higher Mean Time to Resolution
  • Demand forecasting – This is all about capacity waste and stress
  • What are my used vs allocated resources?
  • Do I have the need for burst demand?
  • How can we stay ahead of demand?
  • How can we optimize efficiency and cost?
  • What metrics should I measure?

How do we think differently?

  • Right metrics
  • Change your mindset
  • Look for symptoms

vCenter Operations Management Suite Overview

  • Integrated management
  • Manage physical and virtual
  • Automated management

Is your capacity / performance problem really a capacity / performance problem? Or is it really a configuration problem?

Getting the right metrics

  • Capacity
  • Usable Capacity
  • Demand
  • Usage
  • How much would a VM consume if it did not have contention?
  • Humans can feel about 10% degradation in speed

What are the symptoms?

  • Workload = Demand / Capacity
  • Abnormal = Actual / Expected
  • Faults = Failures
  • Stress = Accumulated Peak Demand / Time
  • Time Remaining
  • Capacity Remaining
  • Waste
  • Density

Be wary of “Faith-based” capacity planning. This is when capacity management tools are tuned based on general best practices and not the day to day run rate of the infrastructure. Tune the management tool based on what an actual “Bad Day” looks like. Ratios are more important than actual capacity numbers.

Looking at workload can help you to quickly identify problems like:

  • A VM could be way overprovisioned
  • Population pressure – is my host overcommitted?
  • Is a limit causing problems?
  • Always check normal demand vs charts shown

Map symptoms to problems

  • Workload High / Anomalies Low / Stress Low – OK
  • Workload High / Anomalies High / Stress Low – Investigate

Putting it all together

  • Get the right metrics
  • Change your mindset
  • Watch for symptoms
  • Putting symptoms together
  • Visuals and notifications
  • Find the problem
  • Finally, remediate

That wraps up this session. It’s good to see VMware thinking about these problems and bringing integrated solutions to the market to address them.

leave a comment