Infrastructure Outsourcing: Part 3 – Bold Solutions Needed From Willing Suppliers


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. In Part 2, we addressed why more virtualization is not the real answer. Where are the next big benefits going to come from and who is willing to make the paradigm shift?

Continuing in our example from Part 2 where our Buyer was looking for $125M in savings over a five year term, if the virtualization dog won’t hunt (well enough) what dog might? Perhaps x86 hardware consolidation should be addressed in a different way in a sourced environment. What if instead of using 15,000 virtual images, applications could be stacked, like they are on other platforms like mainframes. While no application stacking effort would achieve 100% results, neither would virtualization. For simplicity in calculating the virtualization numbers we assumed 100% of the images could be virtualized and we will do so again for the application-stacking alternative. In both cases, what can be achieved in actual implementations will be less.

Let’s assume that each of the 15,000 O/S images runs one application instance. Then let’s take those applications and stack them inside let’s say three O/S images on each of 1,000 machines. We will still need the same amount of hardware, the same amount of virtualization software, which will cost $62.3M over the term, but then let’s stack the 15,000 application images in the resulting 3,000 O/S images. In that case our service fees would drop from $202.5M to $89.1M (15,000 * $225 for 18 months + 3,000 * $225 for 42 months) a projected savings of $113.4M over the term. The $113.4M is 90% of the buyer’s savings goal of $125M.

Perhaps and even probably, the cost of implementing an application stacking approach would be more expensive than $7.5M allotted for virtualization, but the economic benefits of such an approach are over four times better than the virtualization only alternative — there is plenty of room to support a larger initial investment.

To get to a $225 price per image, suppliers have thrown a large portfolio of productivity tools, ITIL processes and low cost labor at the problem. Labor isn’t going to get any cheaper (eventually they run out of yet even lower cost countries); the processes might get marginally better than today’s best practices which may greatly increase productivity (e.g., cloud operators Google, Amazon, etc.) and drive the cost per image to a somewhat lower number, but when can it be expected to be four times lower? When will it get to less than $60 per image per month?

To achieve a true break-through in cost savings requires a radical change. The answer isn’t better, cheaper, faster woodchucks… we need to reduce the amount of wood that needs chucking. There is a huge opportunity for the outsourcing supplier that can offer a cost effective transformational solution to radically reduce Q. Whether someone has the skill, the execution risk appetite and long term market view (accepting lower current client revenues from reduced Qs in exchange for radical growth in market share) remains to be seen.

Mainframes have featured application co-existence and stacking for years. Why can’t the same approach be applied to curtail the viral growth in X86 O/S images and the resulting expense?