A Look at VMware Licensing & Environment Growth

by Bob Plankers on July 14, 2011 · 13 comments

in System Administration,Virtualization

My previous post on VMware licensing changes focused mainly on the question of “will I be able to upgrade my current setup to vSphere 5?” I concluded yes, easily, and if you would like to see how I did it I encourage you to go read the post.

It’s pretty obvious that futures will change, though. I’ve always subscribed to the “fewer bigger machines” theory, to which I owe a lot to Steve Chambers. Some of his writings espoused the idea that most IT failures are human error, to which I agree. Coupled with that is the idea that the cost of a server isn’t in the price, it’s in the management of it. So why have more machines, more network ports, more SAN connections, more of everything to manage and have errors and problems with? Besides, VM performance is better with larger pools of resources, too, especially with many different sizes of VM.

My last strategy, up until Tuesday, July 12, 2011, was to keep buying Dell PowerEdge R810s with dual Intel X6550s and 256 GB of RAM. That’s 16 cores in total. My environment, like many, eats RAM, so I need more of that than I need CPU. So I set out to see how big of an effect this will have on my budget.

Some things I’m thinking of & assuming as I ramble through this:

  1. You can buy vSphere licenses as you go, so you don’t have to do it all at once. However, it’s my intention, when I acquire a server, to fill it up, so I’ll work with that assumption.
  2. A “full” server isn’t 100% allocated RAM, but really 85% according to Gabrie. That covers overhead and such. It might be lower still if you use the HA/Fault Domain Manager capabilities.
  3. For me, growth is triggered when I reach 90% of “full,” so 90% of 85% of 256 is 196 GB. I will use that number in my work.
  4. This work will assume I’m just adding to my already-full cluster, and I don’t need to worry about spare capacity at all, the +1 in the N+1 cluster sizing.
  5. I will use numbers from Dell’s main web site here, as well as list price of VMware licenses. I will discount all prices by 20% to try to keep it more realistic (nobody pays list price, right?). I’ll also assume SnS is 25% of the price of a license, paid yearly.

So, I spec’ed out my baseline R810, with 256 GB of RAM, dual Intel X6650s, dual power supplies, RAID 1 146GB 10K disks on a PERC H700, additional Broadcom 5709 and Intel X520-DA NICs, a 5 year NBD warranty, and “keep your hard drive” for five years. It comes to $22,626. Minus my theoretical 20% discount that’s $18,149, or $6,050 per year for three years if you depreciate.

196 GB of RAM, divided by 48 GB of RAM per CPU, is 4.08 Enterprise Plus licenses, which I will round down to 4. List price for Enterprise Plus is $3,495, but after my 20% discount it’s $2,796, totalling $11,184. SnS per license is $699, so 5 years times 4 licenses times $699.00 = $13,980.

For five years the vSphere 5 total for one of these servers is:

  • $18,149 for hardware.
  • $11,184 for vSphere 5 Enterprise Plus licenses.
  • $13,980 for SnS.

Summing up to $43,313, over five years. That’s $225 per GB of RAM for five years, by the way, or $45 per GB per year.

Doing the same math for vSphere 4, where the major difference is the lack of two more Enterprise Plus licenses, yields $30,732. That is a difference of $12,582 over five years.

If I break it down according to my depreciation schedule, three years for purchases over $5,000, it looks like:

  1. $12,574 vs. $9,312 ($3,262 difference)
  2. $12,574 vs. $9,312 ($3,262 difference)
  3. $12,574 vs. $9,312 ($3,262 difference)
  4. $2,796 vs. $1,398 ($1,398 difference)
  5. $2,796 vs. $1,398 ($1,398 difference)

Can I shave $12,582 out of the cost of a server in five years? Um, no. But $3,262 a year isn’t as big of a chunk, and if I don’t buy the vSphere 5 licenses until I need them that’ll save me money, too. In fact, if I wait until year 3 to buy the licenses I save nearly $3,000. I could also order the hardware with 128 GB of RAM (16 x 8 GB) and get the other 128 GB in year 3, hoping the price comes down.

I have four other thoughts on how to make this budget-neutral.

First, perhaps I can cut down on my support costs, relying on lower levels of support. I actually have BCS support, so for me I’m going to make this budget-neutral in the medium-term by dropping it. I’m not real happy with BCS support, as I don’t ever need help. I just call to report bugs, and they never fix them. If VMware is going to ignore me & my bug reports I might as well go for the cheapest option available.

Second, I could be much more “just-in-time” about growing the environment. That’s a lot of screwing around with it, though. Perhaps the autodeploy features will ease that, but buying licenses is always a big pain. I’d rather just do it once.

Third, I could look at smaller machines. Thing is, I’m not willing to run 800 MHz RAM on Nehalem architectures, so I’d be limited to 96 GB of RAM on Dell R610 & R710s. License-wise, that’s good, but I will lose money in acquiring and managing extra network switch ports, SAN connections, etc. I’d rather just pay for extra licenses for the bigger hosts.

Last, could we frame this as vSphere 5 saving my organization $12,582 in staff time? If we say that an average staff member costs $40/hour to a business when you figure in pay, unemployment, benefits, etc. that’s 314 hours of time, or almost 63 hours of time per year for five years. That’s actually plausible.

Overall, though, this isn’t a welcome license change if you aren’t done growing your VMware environment. It’s potentially very expensive and VMware knows they have us stuck where we have to pay it. I will have to, as the options to save money just aren’t very good. I can only hope that it opens the door wide for competition in this space. Real competition, not the Xens, KVMs, Oracle VMs, and Hyper-Vs we have now.

And yes, I know, this all sounds like a departure from my previous posts. But I hadn’t really worked through all the numbers, so I didn’t want to talk about it until I had. Honestly, I thought there would be some options for cutting costs, but beyond support changes there really isn’t anything that actually saves any money.

As before, I always welcome factual & professional comments and discussion! You can also reach me at @plankers if you find an error or would like a clarification.

Previous post:

Next post: