Are you seeing the real benefits of virtualization

1
Great Expectations

• Are you seeing the real benefits of
  virtualization?

Ron Oglesby Director, Virtualization -
Architecture ServicesGlassHouse Technologies
2
Virtualization is a Strategy not a Project

• Virtualization should be an overall strategy from
  the network, through the servers and applications
  into the storage.
• Virtualization has HUGE expectations, are you
  seeing the benefits?
• Why are some companies missing or not able to
  articulate the benefits of virtualization within
  their environment?
• How can you avoid the pitfalls others are seeing?

3
From a CA study

• the rapid adoption of server virtualization has
  led to heterogeneous environments that are
  proving difficult to manage and nearly impossible
  to assess. CIOs don't have a clear view into how
  their virtualization efforts are working
• Top three keys to successful projects were noted
  as
• Being able to measure performance of the virtual
  environment
• Diligent inventorying of server assets and load
  distribution
• Thorough investigation of available technology
  solutions.

4
What is being said post implementation

• 44 of companies that have deployed server
  virtualization are unsure whether their
  deployments have been successful
• 28 of organizations with virtualization have
  either failed to, or are unsure if they, realized
  a return on investment
• 40 said they have either failed to achieve
  expected cost savings or they simply couldn't
  tell

5
Common issues in virtualization projects

• Unable to quantify the cost savings!
• Failure to virtualize number of targeted servers
  (thus failure to realize savings)
• App owner/Vendor push back
• Monitoring / SLAs / processes integration
• Not enough capacity / performance issues
• The vendor said this would be easy.
• Virtualization should not be taken lightly. It
  should be treated like the major infrastructure
  change or overall strategy that it is.

6
Unable to quantify the savings

• Frequently use fluffy TCO/ROI models
• Vendors assign to downtime and man hours
• Use power in cooling numbers that are high/not
  the customers actual numbers (or IT doesnt own
  power cost)
• Reduced head count frequently figured in, but not
  realized
• Reduced server costs more logical servers more
  mgmt.
• Often do an ROI prior to design and full
  discovery
• ROI should use real and easy to quantify numbers
• Server hardware (first or second year) dont go
  longer?
• Network equipment, rack space, UPS, KVM, cooling
  in tonnage, fiber ports
• Must factor in depreciations

7
Failure to virtualize enough servers

• Failure to schedule the migrations and
  manage/push the scheduling
• Customers/end users not educated on why, when,
  how and benefits they are receiving
• No formal targeted list of servers with logical
  reasoning and priorities
• Depreciated, older servers not targeted
  aggressively enough, and new servers ignored
• Not enough resources (qualified resources) to
  migrate AND manage growth/change and integration
• Lack of Mgmt buy-in / Push not viewed as a
  major change Again strategy, not project.

8
App Owner/Vendor Push Back

• Common push back messages
• I need four Processors and 8 GB of Mem
• We dont support that application on a VM
• My application doesnt work on a VM
• Response / Mitigation
• Have real data on every server (workload /
  utilization)
• Of the thousands of x86 servers we have looked
  at 81 use less than 500 MHz Peak hour, 92 less
  than 1GHz peak hour.
• Have a solid back out process as mitigation, VM
  to Phys and workload migrations should not be
  limited to host to host.
• Of all the servers we have done only one ever had
  issues as a VM

9
Monitoring and Integration

• Virtual Center is not the end all of monitoring
  and does not do capacity planning
• Integrate (from day one) with existing monitoring
  and workflow tools
• Change and configuration mgmt is important
• Third party tools for integration with HP
  Openview, MOM, etc. are available, use them!
• Integrate into existing build / decommissioning
  process

10
Capacity / Performance Issues

• Initial irrational exuberance VMs for
  everyone!
• VM environments often overrun their initially
  purchased capacity (cost and perf issues)
• Virtualization makes hardware cheap so it gets
  used quickly (VMs arent free, more on that
  later)
• Performance issues often come from inexperience
  in candidate selection and/or improper configs.
  Anything can be virtualized, but at what cost?
• Capacity is a VM thing, not a processor / memory
  / specific metric.
• A capacity plan is ESSENTIAL!

11
Capacity, Capacity, Capacity

• Its about the number of VMs you can deploy, and
  how fast you will burn that up
• Over utilizing/under estimating use can
• Cause poor performance
• Bad impressions with app owners
• Long term negative views of the technology

12
VMs are not free - have a cost model

• Free VM environments will run out of capacity
  fast (free more usage)
• Used capacity needs to be replaced
• for todays VM can buy more servers
• Most organizations dont implement a cost model
  because they
• Cant quantify the cost/dont want to/its
  somewhat complicated
• Afraid of implementing charge back but it
  doesnt have to be
• See it (the funds) all as one big bucket and
  ignore how servers are paid for now

13
The vendor said it would be easy

• I love the software vendors, but they all want to
  sell licenses
• Like most software, hypervisors install easily.
  It is easy to click next
• Imagine SMS or Altiris or Openview being deployed
  in your environment sure you can click Next
  but do you that? Or do you Design, Plan and
  Deploy like the major technology it is
• Vendors are right Its easy to install but,
• Its not easy to get right as a strategy

14
Final Tips

• Full discovery and rationalization prior to
  project is key, this feeds the ROI, Design and
  Planning
• Look at numerous alternatives (for software and
  hardware) and define design based on requirements
• Design strategically! Then act tactically.
• Use a solid ROI model and validate it constantly
  during the project Get a VM COST MODEL
• Define the critical success factors, make them
  known, and measure the project against them
  constantly
• Integrate it into all existing tools and
  processes before the first VM is deployed

15
Thank you Questions?

• Ron Oglesby
• Director, Virtualization - Architecture Services
• www.glasshouse.com
• Ron.oglesby_at_glasshouse.com