Infrastructure Outsourcing: Virtualization Doesn’t Solve the Problem (Part 2 of 3)


In Part 1 of this blog post (Time To Mind Your Ps and Qs), we made the case that there is limited additional opportunity in continuing to pound on “P” in the P x Q = Total Price equation, and that to achieve the next breakthrough the supplier community has to address “Q”. The current standard answer from suppliers on reducing Q is “virtualization”, but that won’t solve the problem, at least not entirely. Here’s why.

Assume we have a buyer with significant IT Infrastructure labor costs — say $125M per year. The buyer decides to go to market despite having a pretty good idea that its unit costs are roughly at parity with current managed services market pricing. The buyer’s objectives include, in addition to qualitative and risk mitigation goals, lopping $20M to $25M p.a. off the labor costs to manage its infrastructure. A five-year labor-only deal in the $500M TCV range is certainly going to attract plenty of interest in today’s marketplace. The buyer has made a strategic decision not to source hardware and software ownership to the supplier so, if necessary, they can “fire the maid without selling the house.” Furthermore, the buyer has decided to signal to the suppliers that its unit costs are near where it believes the market should be and winning this business is probably going to require a clever solution that addresses the Qs along with the Ps.

So, let’s first look at this from the supplier’s perspective. If you are the clever solution developer at a major supplier, you see a way out of this conundrum. You’ll propose a virtualization initiative for the buyer’s vast portfolio of x86 servers! And, since x86 services are typically priced by O/S images, you will still get the same amount of revenue regardless of the degree of virtualization, 15,000 images on 15,000 machines or 15,000 images on 1,000 servers — all the same to you, right? However, since this is a labor only deal and you will be reducing the quantities of something that isn’t in your scope, you have to change the way the buyer calculates benefits to include all the ancillary stuff they won’t buy from you anyway (i.e., floor space, energy, racks and, other than a couple of suppliers, the machines themselves). Starting right in the executive summary you will tell the buyer to think strategically, not tactically. That is, think about TCO, not just about this isolated deal when calculating benefits. You are still going to have to employ a lot of “weasel words” to deal with how virtualization will occur (and how fast) — but at least there’s a story to tell.

But from the buyer’s perspective, how good a story is it? Let’s assume the buyer uses a standard, one image, x86 device that costs $3,600 to purchase and the buyer depreciates that device over 36 months. That is $100 per month in depreciation. Let’s ignore the per-image software costs and just assume that it will cost the same if it is running in a virtual or a real machine. Then lets assume that a device big enough to run fifteen virtual machines will cost $15,000 or a $1,000 of hardware cost per virtual image. That’s a real good start because the buyer just reduced their hardware costs by $2,600 per O/S image. With15,000 images, the buyer just “saved” $39M over three years and $65M over the five year term. Doesn’t that look compelling, right there at the beginning of the executive summary? The buyer is looking for $125M over five years and with a little bit of virtualization, a good 52% of that is covered with this one initiative.
Well … hold on just a second; it’s not that simple. You have to do transition before you can do transformation, and that is going to take up to about a year, and the transformational initiative takes time too, so lets say you are only going to get 3.5 years of the full benefit stream not the five years you initially thought. So the benefit drops back to about $45M.

Then there is the virtualization software. Let’s assume the buyer can get the software for about $7,500 per server and has 1,000 servers. That’s $7.5M in software. The buyer needs to pay maintenance on the software (let’s assume 15% p.a., with a one year warranty, so with maintenance for 2.5 years), that’s another $2.8M. So the virtualization software is $10.3M over the term. Now the projected benefits have dropped all the way back to $34.8M, but you haven’t yet included any costs for actually doing the initiative.
Now, let’s assume the Buyer is pretty good at doing these and they can convert an image from real machine to a virtual machine in four hours and you have SME’s who can do the work available for $125 per hour. For 15,000 images that is another $7.5M in cost to the buyer. The net benefit is now $27.3M or about 22% of the buyer’s goal. If you do all of the arithmetic it turns out the entire exercise would save about $43 per month per image. Interesting, but it fails to meet the buyer’s financial goals. And the whole initiative ignores where the real money is – the $125M in annual labor costs the buyer is looking to reduce because there are still 15,000 images to manage.

Well what about others things like floor space and energy? Energy and floor space consumption will go down, but this IT buyer, like most, doesn’t pay for real estate or energy, someone else in the enterprise does — so the benefits will accrue elsewhere and not show up in the business case except maybe as a footnote. (Not to mention that real estate is hard to get rid of, especially in down markets, so freed up space often just sits there empty in the short term.) This is clearly wrong-headed from the shareholders perspective, but a fact of life in most large enterprises.

In a five-year deal, a good guess at what the supplier might offer on a per O/S image basis might be around $225 per month, per server and all of the work and change activity will have done nothing to address that number. In a sourced environment, the x86 Q that counts is O/S images, not machines, not virtualization ratios — just O/S images. Doing the arithmetic one more time shows that over the proposed five years the total cost of the virtualization initiative is about $63M ($45M for hardware and $18M for software).

Compare that $62.3 to the projected services price on 15,000 images of $202.5M. Suppliers may want you to focus on the value from virtualization, but if you are looking for substantial reductions in ongoing delivery costs, it won’t get you there because it doesn’t attack the amount of work that must be done to manage the images on the virtual machines.

Certainly virtualization offers operational and availability benefits as well as some savings, but it’s not the transformational leap clients are looking for.

Next: Part 3 – Bold Solutions Needed From Willing Suppliers