Cloud ROI, 5 areas of returns

[#1]

In many conversations the question on Cloud return on investment pops up. Many articles have been written on the subject, each taking a different approach. In my mind there are the tangible benefits, which I will focus on today, and the intangible ones such as agility, responsiveness, speed of provisioning of infrastructure/applications etc.

Server Efficiency

It’s widely agreed upon that dedicated servers have an average efficiency (time they are really used) in the area of 10 to 15%. I have tested these numbers with many CIO’s and most agreed on those numbers. This means that, at least 85% of the time, they consume energy, take space and need management without actually doing anything for the company. Decoupling the application from its hardware through virtualization and allowing the environment to be used for more than one application improve the efficiency of the server base. The capacity-utilization curve created by Amazon illustrates this well, although it only looks at the increase in requirements, which is obviously not always the case.

In an article, titled “Clearing the Air on Cloud Computing,” McKinsey states that through standardization and aggressive virtualization, efficiencies of anywhere between 30 and 35% can be reached. This means that one could do with only half the servers, implying a 50% reduction in energy consumption, in maintenance contracts, in space requirements. It also implies less operational resources are required to manage the environment, and obviously the capital expenditure is also lower.

In a detailed study, VMWare points to following results:

• ROI over first six months: 189%
• Reduction in hardware and software costs: 63%
• Reduction in IT operations costs: 70%
• Reduction in procurement costs: 50%
• Average CPU utilization: 5–10% before virtualization, 60-70% after
• Server consolidation ratio achieved: 12:1

My only comment would be that CPU utilization from 60-70% can only be achieved in companies having a very stable IT demand. But even if we stick to our 30-35% and reduce VMWare’s numbers by two, we have savings between 35 and 40% of costs, leading to ROI’s within the year.

Storage Efficiency

How much storage do you allocate to users and applications? How much of that is actually used at any moment in time? It’s probably less than 50%, so if we virtualize storage, removing the direct link between the storage allocated and the actual storage available, we do not need a total storage corresponding to the sum of the allocated storage. This is what storage virtualization is all about and what companies such as 3PAR have become specialists in. To just give you one example, a UK based Fund Management company claims 90% reduction in provisioning time and 50% reduction in maintenance costs, not talking about the reduction in capital expenditure.

Operation Staff Reduction

Complementing virtualization with automation addresses another level of cost. Indeed through the automation of many key IT processes, not only does one need fewer operators, but mistakes are also diminished. And you remember Amazon’s latest outage; it seemed due to a human error. Amazon actually stated "we will audit our change process and increase the automation to prevent this mistake from happening in the future.”

HP performed an in-depth study on the benefits from automation, looking at 9 areas where automation brought benefits. I just want to focus on the staff reduction, here are their conclusions:

• Storage administrator ratio: Doubled the amount of storage with the same number of administrators after deploying HP Storage Essentials.
• Desktop system administrator ratio: Cut the number of system administrators per 1000 desktops by a factor of 10.
• Server system administrator ratio: Ratio grown from 1:100 to 1:150 administrators per servers.

In other words, staff was reduced by at least 50%. Now those numbers are somewhat misleading as one would have to add resources to create the automation workflows in the first place, so in the short term the benefits may not be that great, but once the automation remains in place it can be used over and over again.

Call Center/Helpdesk costs

You remember the time you needed to call to request a new service? Well with self-provisioning that is over now. The provisioning calls no longer require call center resources to address them. You actually do their role by choosing and requesting the application on the portal. Although you may argue such savings are smaller, they should definitely be included in the mix.

Licensing

Where the four first areas I describe are easy to understand and to calculate the results. There is actually a fifth one, and that is the one associated with the license cost of your key applications. This time, everything depends on the type of license you have. Obviously if you own an enterprise wide license, the fact you take and release license keys thanks to cloud won’t matter, but if you have a seat license for example, the careful management of the provisioning and release of cloud services may result in a better usage of your existing seats, avoiding having to buy additional ones. Unfortunately, ISV’s only get to grip slowly with the concept of cloud and a pay-per-use approach to software licensing. That’s why, although better management of existing license schemes can already take place in the cloud, the full benefits will only be reached when ISV’s change their licensing approaches.

And what about using the public cloud?

In the current post I highlighted savings based on the use of private cloud environments. Is the public cloud actually cheaper? That depends. We studied Amazon and Google cloud service pricings and when capacity is needed for periods of less than 8 weeks per year, public cloud options are actually cheaper. For longer periods it actually hints the other way. McKinsey, in the study quoted earlier, came to the same conclusion. But this brings us to another point, would it be possible to combine private and public cloud to optimize ROI? That’ll be for another blog entry.