2

As a virtual infrastucture admin, what metrics would you tend to be interested in to deduce whether a server virtualisation project rollout is successful/saving the company money? So whether it saves the company money over the purely physical approach, etc.

So the metrics need to be giving business information, as opposed to technical information (e.g. memory % used).

Thanks

GurdeepS
  • 1,646
  • 5
  • 26
  • 33

2 Answers2

2

I'm thinking ROI calculation, which is otherwise known as a cost benefit analysis. Off of the top of my head, here are some areas to consider...

  • Power savings
  • Server equipment savings
    • Need to consider big expensive server versus multiple small servers, often the ROI is minimal here.
  • License savings
  • Infrastructure savings
    • UPS
    • Cooling
    • Racks, switches, KVMs, cables, PDU, et cetera and so forth.
    • Lease/buildout, physical space cost
  • Time/people savings
    • Time for provisioning a new server
    • x less staff necessary for running y servers

I recently had a cost benefit analysis performed on virtualization, which ended up being more expensive than my current infrastructure. Commodity servers and an Open Source software stack is difficult to beat on cost. Storage and licensing can be expensive for some virtualization solutions.

Community
  • 1
Warner
  • 23,756
  • 2
  • 59
  • 69
  • Power is the big one here. Switching from individual servers to a fully-loaded Dell 1855 Blade Centre with vSphere dropped power consumption by about 30%. Switching from 1855's to 1955's dropped power consumption by an additional 65%(!!). Our power bill is now down to $1000/quarter. – Mark Henderson Jul 20 '10 at 23:01
  • Great lists. Good inspiration! – GurdeepS Jul 21 '10 at 21:59
1

A good metric would be the number of physical machines you saved in the virtualization process, if you are providing 90 vms with 10 servers, you saved a lot of money in hardware by better using the available resources. Less machines also mean less energy cost, less space and cooling.

Speed in setting up new servers can also be a good metric. How much time you used to took while creating a new server (unboxing, formatting, installing, etc...) and how much time it takes now.

If you have migration strategies working, the time you take to bring servers back up after a hardware crash is also a factor.

Those are some I can think about right now.

coredump
  • 12,713
  • 2
  • 36
  • 56
  • 1
    I don't know that you save that much on hardware, considering you can get ~5-10 "small" servers for the price of 1 LARGE server. Also realistically it takes about 10 mins to unbox and mount modern servers with rapid rail systems. Formatting/installing should be done by imaging already not sure how much time you would save there. Recovery time from hardware failure is a good thing to have compared in your metric. – Zypher Jul 21 '10 at 04:56
  • I'd say you save a bunch, if your VM:host ratio is acceptable. We're running 80VMs per host in one environment and 40 in our core environment - Those kinds of consolidation ratios equal a massive hardware saving. Factor in the cost of spare parts/maintenance/ilo licenses and you're looking pretty sweet, even factoring in VMWare licensing costs & maintenance. Unboxing etc includes switch port configs, power provision, initial network config. Soon adds up. – Chris Thorpe Jul 21 '10 at 05:46
  • You are spot on. Things like migration time are definately big factors. – GurdeepS Jul 21 '10 at 22:00