The Data Center Dilemma: Build or Outsource?
May 14, 2013, Uptime Symposium, Santa Clara, CA—Kevin Brown from Schneider-Electric proposed some evaluation criteria for supporting a make or buy decision for new data center capacity. The primary pieces for evaluation are cost, cash flow, and strategic factors for upgrade, build additional capacity, or outsource.
Existing evaluations are based on rigid financial models and common financial ratios. Instead, a set of simple creation and evaluation models can provide good approximations and incorporate strategic factors with a structured approach.
A better financial model uses financial metrics that are normalized for a 10-year total cost of operations per watt. It uses cumulative cash flow and quickly identifies break even or crossover for operating and capital expenditures. The idea that a simple model is sufficient for quick iterations and is close enough to identify major benefits and drawbacks is fairly unique. The caveat is that it is not to the same level of detail as a full budget.
For demonstration purposes, some assumptions for total cost of ownership for a 10-year timeframe might be $62/W with rates increasing 5 percent a year. A remote worker is $225/hour for 1 hour per month, and a 3 percent TCO discount rate. The decision process would be to define there strategies; evaluate the existing plant and clean it up, increase server density, or add capacity with new facilities. If you outsource, the costs of hardware and other facilities is zero, and in included in the costs for the services. See more details in the white paper at http://www.apcmedia.com/salestools/VAVR-8HF8DC/VAVR-8HF8DC_R1_EN.pdf In addition to the whitepaper, the company has an interactive tool for evaluating build versus buy options at http://www.apcmedia.com/salestools/WTOL-8ZVREZ/WTOL-8ZVREZ_R1_EN.swf
The first scenario is to pick the low hanging fruit. If we assume a TCO of $17/W plus $2.10/W for the upgrade. The costs of this effort are much less than outsourcing. A second scenario is to increase density. The 10-year TCO is $32/W plus an additional $4.80 for equipment. Again, the total costs are less than the outsourced $62/W
By adding to capacity with new facilities, the 10-year TCO is $27/W with upgrade costs of $8.75/W. Here, the cross-over for new versus outsourced is about2 years. Here the new focus is on cash and cash flow.
If the plan is to build a completely new data center, the key assumptions for the TCO numbers are an issue, but the financial people will want to see the impact of the cost of capital. The simplified model is sensitive to capital costs for TCO. So as cost of capital increases, the move to co-located facilities becomes more attractive. At a 26 percent rate, the TCO for building and outsourcing are equal.
In addition to the costs, another area where the decision might be influenced includes the organization’s “soft” issues. These include the company culture, the comfort of the IT manager with housing equipment in shared space, IT access to the data center, suspected onerous contract terms, and critical facilities expertise, especially for energy management. More complex issues include regulatory issues, life expectancies of the data and the equipment, deployment timeframe, total capital expenditures and timing, and total cash flow. Different companies will have different issues, and any of these might be critical to the company and its data operations.
The decision makers need to take the time and get the assumptions and soft issue lists to a manageable state. Once this process is firmed up, then the various affected parties can make quick what-if iterations and have a common framework for the overall evaluations and decisions.
The market forces are driving upper layer resiliency and reducing costs. Co-location providers are making system-level and architectural changes that are reducing the costs for enterprise data centers. The co-location model is moving towards a modular data center, and the concept of a much more dynamic data center. As these changes take place, the reliability is increasing, to the point where some of the cost assumptions for a tier 4 system can be displaced by a lower level system at a greatly reduced TCO.
The changes in systems is becoming documented in the open design standards. The co-location systems are reducing the cost of all designs with reference designs and standards. One result is that prefabricated builds are more realistic. A partial build is probably optimal, since the cost of square footage is cheap, and the availability of an existing area allows for incremental upgrades.
A tiered data center can be set up for tiered apps. The mission critical apps go on a tier 4 system, and lower level apps can reasonably operate in a tier 1 or 2 level of reliability. This approach provides an opportunity for facilities and IT to talk, review standards, and assess the underlying assumptions. The process brings about coherency of the various business models.
