Data Center Buy vs. Build: Preliminary Considerations

Editor's Note:  Author Tad Davies is a 28-year veteran of the data center industry and a member of AFCOM's Data Center Institute. Currently, Tad leads Bick's Consulting Services business unit based in St. Louis, MO. He will be presenting "Data Center Construction 101 for IT People Only" along with DCI co-board member Bill Doty on Monday, April 3 from 8:30 a.m. to 11:30 a.m. at Data Center World Global - 2017 in Los Angeles. Register today! This  previously published AFCOM article is just a sample of his expertise.

By Tad Davies

There are innumerable writings addressing data center build vs. buy. Unfortunately, service providers with an agenda are producing most of them. Here is some straight talk based upon strategy work with hundreds of data center owners. 

Time is a commodity that is increasingly less available. If you have been assigned the task of analyzing the merits of build vs. buy, most likely you have not been given a reprieve of your current responsibilities. Working efficiently will save you dozens to hundreds of hours depending on the size of your project.

Understanding the ecosystem and gaining alignment around requirements is a critical first step. First, let's consider certainties as pertains to this endeavor. I refer to them as Davies' Dictums.

Everything is Situational

Every organization has different priorities and they change over time as a function of various aspects including board/executive concerns, business focus, and competitive conditions. Organizations have different opinions and financial thresholds involving Capex and Opex. So it is logical that no one solution is best.
Things change on the supplier side as well. Space availability changes as contracts are signed and lapse.

Think of purchasing colo space somewhat like airline seats. The rate that you expected or were quoted months ago may not be valid, which is certainly understandable. Many providers are offering a wider variety of services, blurring the lines between colo, managed services, and cloud. On top of this, providers are buying others or re-aligning their portfolios.

Building vs. buying is ultimately a matter of understanding your business priorities and assessing the market's current offerings. After the facts are defined, you can make a decision about which course is best for your organization. To illustrate, here is a point/counterpoint view of building versus buying:

"Building my own data center is the best option"

Pros

Install custom solution options I want

Size infrastructure according to my needs

We have capex to invest

I can manage my own facility

My costs will be less than theirs

I have operations and technical staff

Cons

You likely can't afford all the a 3rd party can. Most 3rd party DC's have more robust infrastructure

You will have to significantly "over-buy" facility infrastructure. Can't go back to the board in 3 - 5 years asking for another capex infusion

Are there better uses? Plant, equipment, IT infrastructure?

Operational procedures in most 3rd party DCs are excellent. Their business model depends on it

Sometimes not all costs are included in your analysis. e.g. real estate, maintenance, licensing, communications, taxes, etc.

3rd party may have more proficiency and technical expertise that you can "rent" to augment your staff or provide staff redundancy. Let's say your SharePoint person leaves.


"Colo/managed services is the best option"

Pros

Colo has better infrastructure

It will increase availability

Provide capacity increases as needed

Avoid major capex event

Enable organization to focus on strategic vs. tactical IT

Leverage provider's technical expertise

Leverage operational processes - battled hardened and field tested

Cons

Current owned data center infrastructure has a proven track record and has plenty of "life"

No history of outages at current site

Technology has shrunk IT infrastructure so there is plenty of capacity at the current site

Capex is not a concern but opex is (e.g. universities). Cost of $ is low now

IT is the core function of the business. Tactical experience required

Expertise resides in house and is available now. Besides it's custom

Proven processes and track record already in place


What Matters to Your Organization is What Matters
While this statement seems obvious, I have found many organizations get caught up in the arms race of features/services amongst providers. If the feature set does not impact your organization today, or possibly in the future, then it should neither favor nor detract from your ranking of that service provider. Forget it and move on.
Providers also fall into this trap by blasting out features before discovering what matters to the organization.

Bad News Early is Good News

If something will affect the outcome of your decision making, finding out about it earlier in the process will save you time, frustration, and money. Here are some examples:

  • Executive/board preferences eliminate the option you have been pursuing. Float trial balloons early, and get a coach who has access to the C-suite.
  • Technical limitations (such as application latency). Have the appropriate SMEs involved in creating the solution.
  • Financial implications that are not defined early. Submit preliminary budgeting to ensure that your request is financially viable. And remember the situational dictum - what was once acceptable may no longer be.

This dictum applies to providers as well. While hard to stomach, if you know that a provider(s) is not a fit, best to let them know early. It is unproductive for you and the provider to continue the conversation. I have found retail colos trying to be included in a wholesale colo RFP. Like taking off a Band-Aid, doing it right will hurt at first.

What are You Buying? Your team will have to be aligned regarding what services you will consider procuring. Are you considering colocation only? Managed services? Cloud? Maybe a combination (hybrid)? Finalizing on a requirement will simplify your criteria.

As you consider how best to meet your needs, you may come to the conclusion that one size does not fit all. You may need to consider various services that require more than one provider. This activity will consequently inform with whom you will eventually solicit RFPs.

What are You Currently Spending?

If you do not know what your organization is spending to support having a data center in-house, how can you make fact-based decisions on the financial merits of a third-party service? Conduct a Total Cost of Ownership analysis and segment your analysis by the services you may consider procuring. Colocation is the most straight forward and should include cost components such as:

Utility - electricity, water, etc.

Building/grounds maintenance

Consumables (annual)

Parts

Communications

Equipment maintenance

Non-recurring consumables (UPS batteries/fans capacitors

Depreciation

Space - rent

Equipment break/fix

Facilities staff


An analysis comparing a managed services provider to your cost of business becomes more complex. You must determine what specific services you will procure. In addition to the above costs, you might include IT infrastructure and the staff that supports the following functions:

IT infrastructure

OS patching and support

Security/Firewall

Monitoring

Middleware support

ERP (SAP, Oracle, Infor)

Capacity Management

Support apps (Exchange, SharePoint, etc.)

Disaster recovery

Request that someone from finance participate in this analysis. That person knows how to assemble a business case specific to your organization, where to get the data, and how it should be presented to executives.

The term of your analysis will be a function of your business needs but should typically be 10 years for a colo and two or three, 3-year terms for a managed service provider. You will have to take into account replacement of assets at your data center if any are approaching end of life.

What are Your Capacity Requirements?

You will need to build a multi-year model reflecting how much infrastructure will be required to support your requirements. Typically you will need to provide quantity of racks (or rack equivalent units- REUs) and power consumption per rack. The provider will want to know the maximum power requirement for a rack. Consider providing some level of detail such as a kW value and quantity for server, storage, and communications racks. Assign a growth factor and include a density factor as that will likely increase over time.

Get your SMEs involved in this exercise. They should be held accountable making the model reflect reality. While it is impossible to predict the future, state all of your assumptions to provide an audit trail in the future. The key thing to remember is that this is a model to enable comparative decision-making. While it must be representative, you can tweak it later. Do not get caught up in analysis paralysis. The model will also apply to the option of building a new data center or expanding an existing one.