Slow Down to Speed Up: 3 key steps before you build, colo or cloud
Before my team can provide clear guidance on the sizing of a data center (whether build, colo or cloud), a lot of front-end discovery must occur before we begin a new data center design. Unfortunately, the data center industry is guilty of not taking the time to “slow down” before we “speed up.” It’s the results of a highly transactional, highly dynamic industry that we work within, that everyone needs it “yesterday.” This inability for our industry to “slow down” is starting to show in several data centers we walk through, typically resulting in oversized mechanical, UPS, and generator systems. Under sizing equipment can be as equally painful, but gets overlooked more so because, “the organization is growing” or so the story goes.
As a mechanical engineer, I’m eager to design a data center and assess how it can be effectively cooled. In some cases, data center experts forget to allow time for our IT counterparts to inform the design that drives the overall success of the Data Center. Same can be said for electrical engineers, contractors, vendors and others that claim the “data center expert” title.
Engineers work hard to understand the key question of, “how much power do you need”? We often resort to looking at the nameplate of the current UPS, plus some estimated growth factor. Some consultants go a little deeper and study the IT load within the racks, however most high-level power studies can be amazingly insufficient. If a customer is determining their colocation needs, a true understanding of the footprint and power is the key requirement to the request for proposal, and all too often, grossly misestimated. We need to dig deeper and better understand the IT load – i.e. the servers, storage, and network equipment in the racks and how they are being utilized.
At the end of the day, the true goal is to bring together facilities, IT and the executive suite to right size the Data Center, and ensure it meets each team’s goals and objectives. This is no easy endeavor, but requires our team to “slow down” to allow customers to “speed up”.
Here’s a high-level review of the three steps we take before reaching that all important question, “How much power do you need”?
Understanding the corporate goals
Our first and foremost step is to determine the purpose, mission, and vision of the organization. Lack of understanding of a company’s mission, vision or goals can lead to missed expectations, and missed goals and objectives. Also understanding if the company is expanding, merging, acquiring other companies or organizations, is a key goal to understanding of the trajectory of the organization. All roads do eventually point back to the data center. Understanding the company’s purpose allows us to understand how the corporation will leverage and use the data center as one of their critical tools.
Understanding IT Requirements
Our next step is working with IT to understand the assets in the data center that are supporting the corporation vision, mission and goals. Those assets are where most firms fall short and head right to step 3. This is why you see data centers in low-load conditions, wet-stacked generators, temperature control not within ASHRAE TC 9.9 conditions, undersized UPS situations, reliability gaps and overall operational nightmares. Not understanding the current utilization is one problem, but not knowing what IT is planning for in the next 3-5 years for servers, storage, network is another.
Current utilization, while on the surface should be easy to account for can be surprisingly over-estimated. This is due to aged equipment running in the data center that have yet to be migrated and decommissioned onto new platforms. This can lead to overestimating the current utilization level of a data center.
Future utilization can be equally, if not more difficult to forecast. Once again, this research starts at step #1 in order to understand how the company plans to grow in the next five years. It is our responsibility to ask the right questions to determine what assets will be required to make that growth feasible. Studying, analyzing and assessing the IT forecast is a crucial step to any potential data center build-out and one step that unfortunately are commonly missed due to external pressures or condensed timelines.
Now that we know the IT assets, we can help determine the consequences of the “what if” scenario. “What if the data center goes down?” and who is impacted. Can the company still produce widgets, or process claims, or sell more tickets on their website? Therefore, understanding the consequences, should the room have a failure, allows us to properly plan for ways to mitigate risk for the organization. Understanding the demanded reliability requirement for the facility can then be quantified, which leads us to step 3.
Understanding the facility (Data Center) requirements
Now that the team has established the proper IT growth chart and projection (per step #2), it gets us to that all important question of “how much power do you need”? This allows the Data Center M&E team to properly size the supporting mechanical and electrical equipment. It also allows us to understand the proper footprint of the facility to fit the IT assets and future growth. While we realize the IT forecast can be limited, the day 1 load assessment helps set the data center up for success.
However, one key step remains. By compiling the data gathered in steps #1 and #2, and understanding the power requirement, the team can then place a value on specific needs for reliability. The math of how to calculate this “cost of downtime” is an important exercise for all organizations to consider. Too many times, we see data centers requiring a specific level of reliability, and the data center is not matched to those requirements. It’s an interesting “Ah-Ha” moment when we start talking in the terms of consequences should a data center go dark. Then we see the CIO, the IT team and the Facility team recognize the weak links in the data center. These discussions help our team formulate the proper level of reliability demanded by the organization. At that point, we are starting to get the clearer picture on footprint size, power requirement which allows us to begin conceptual design of the mechanical and electrical system and provide a meaningful layout of the data center. This applies for colo, build or cloud.
“Slow down to speed up” should be our mantra when making any decisions regarding data center strategy. Whether build, colo or cloud, we always recommend starting with a holistic view of the organization, diving into the applications and IT platforms, and then moving into the reliability need and the sizing of the M&E infrastructure. Even on the most tactical of projects, you should have a data center roadmap. Slowing down could be the best decision any organization makes in order to allow them to speed up for the future.
Timothy A. Kittila, P.E. –
Tim’s responsibility is to provide pre-sales support for the design team and project manager as needed. Tim is a resource for pre-sales activities and project budgeting support. Tim will help prepare the design documents, provide detailed financial estimates based upon preliminary suggested designs, independent of vendor or equipment selection. The design lead and project manager will assist on finding the best “tool for the job”, instead of picking a tool and making it fit. Tim has provided similar efforts for major Twin Cities’ data centers for nearly 12 years.