by David Appelbaum

Data center IT professionals today have singled out capacity — that is, power, space, compute (including storage and network) — as one of the top data center challenges. According to recent research, the surprising findings reveal the most critical problem facing today’s data center isn’t about having too little capacity to handle data center workloads; it’s having too much capacity that has not been properly allocated.

There is effectively a gap between how much capacity a data center is equipped for and how much capacity is ultimately available for productive use. This capacity gap can affect a facility for its entire useful life of 10+ years. Not only does the capacity gap prevent the data center from functioning at peak performance, it also costs IT significant amounts of capital budget for hard assets (such as servers, switches, storage) and expensive resources (e.g. power) as IT attempts to overcome this perceived lack of capacity — to the point of building entirely new facilities to handle the processing load.

The Roots of the Capacity Gap
The problem starts during the initial data center planning phase, where a number of assumptions regarding future requirements have to be addressed before a single brick is laid — let alone a server racked.

The current technique for data center capacity sizing is a brute force method of allocating maximum capacity from the outset. This technique minimizes the impact of unknown future technology and load requirements that could prevent the data center from meeting its processing SLAs.

The cost of insufficient capacity is very high and has to be eliminated as a risk factor. It is extremely costly to increase capacity incrementally through the data center lifecycle without also increasing the risk of downtime. The load requirement of the datacenter will certainly increase during its lifecycle but, as with technology, it can’t be reliably predicted.

Given these constraints, IT’s tendency has been to massively over-provision the facility rather than risk outages. The ramifications, as IT is discovering today, have huge cost and efficiency implications for the data center management group.

In a study of the data center lifecycle commissioned by a leading vendor of heating, ventilation, and air conditioning/Building Management System (HVAC/BMS), it was discovered that the installed infrastructure capacity is equal to the data center’s planned capacity from the very beginning — the facility starts operations fully built out.

The typical utilization plan calls for the expected load of the data center to begin at 25 percent of capacity and that the ultimate expected load is typically estimated to be 80 to 90 percent of the planned capacity (which includes a reasonable 10 percent reserve for peak demand headroom).

However, in real world usage, the actual start-up load is typically at 20 percent of the ultimate design capacity and the actual ultimate load reaches only about 60 percent of planned capacity! Even with a 10 percent reserve capacity, this leaves an incredible 30 percent of capacity unallocated and unused — the capacity gap.

This capacity gap impacts data center costs in two very specific areas: capital costs and operating costs.

Excess capital costs include the costs of the power and HVAC equipment and the associated design and installation costs (wiring, ductwork, ventilation, etc.). In a typical 100 kW data center, the power and cooling equipment can have a capital cost of more than $870,000, or $8.70 per watt. Based on the capacity gap analysis, 30 percent or $261,000 of this investment is wasted — as the majority of data centers are 100 percent built out the day they open.

In a data center’s early years, this waste is even greater as the actual load can be up to 5 times less than the planned load. If we include the amount of estimated interest earned on the original capital investment for infrastructure, the loss due to the capacity gap can be close to 40 percent of the entire capital cost!

Excess operating expenses must also be included in the costs associated with the capacity gap. These include the data center’s operating expenses and asset maintenance contracts as well as power consumption costs.

Annual maintenance is typically pegged at 10 percent of the asset cost, so over the lifetime of an asset, the total cost of ownership can be almost as much as the initial capital cost of the data center equipment itself, if maintained per manufacturer’s instructions.

Excess electricity costs are also significant if data centers are underutilized, and can be as much as 3-5 percent of its overall power rating. Factor in cooling, and it can easily reach 10 percent.

The real cost of the capacity gap is in the perceived need to bring more data center space online to meet the actual processing requirements of the business. This problem daisy-chains down, as the next facility built to bring the additional required capacity online must deal with the same capacity gap as its predecessor. As such, this requires an additional facility to meet its full processing load, which in turn starts the process all over again.

The bottom line is that data center capacity is planned extremely conservatively on the high end of the utilization spectrum, so as to meet the needs of unknown future technology and unpredictable demand. This, combined with ineffective data center capacity management tools, creates a situation where the capacity gap is not just a byproduct of poor visibility, but almost a preferred (albeit costly) outcome to minimize service interruption and downtime risk.

The Management of Capacity
By actively managing data center capacity — power, compute, network, and space — organizations can ensure balanced data center utilization and performance optimization, while maintaining reasonable cost controls.

While many data center tools have been available for several years, few of them have moved beyond the facilities management functions of space, cooling, and power management. Most of these tools rely solely on vendor nameplate values for estimating capacity consumption, which create an inaccurate picture as to how the asset will function under real-world loads and utilization. This forces data center personnel to equate an erroneous peak load measure with a real average load and thereby create overinflated capacity base lines.

An additional challenge is a lack of long-term trend data of utilization patterns and capacity consumption. While most of these systems provide real-time capacity alerting and monitoring, almost none of them provide time series analysis. Understanding utilization, load, and capacity trends are critical for accurate predictive analysis and for establishing a more realistic capacity baseline for an enterprise based on the unique rhythm of the business.

The Future of Capacity
The next wave of data center solutions are now reaching across the aisle to integrate critical IT capacity, such as compute, networking, and storage, to provide holistic data center capacity analysis and what-if scenarios to test new hardware and deployment strategies. These solutions also provide deep asset visibility to distribute load more effectively across the entire facility.

The bottom line is that data center infrastructure must become more elastic to ensure agility in meeting business requirements and managing costs. This must start at the beginning. Real data center capacity visibility leads to correct capacity provisioning, which in turn eliminates the capacity gap and leads to a more efficient and effective data center.

David Appelbaum is the vice president of marketing at Sentilla Corporation.

by Dominic Monkhouse

As the importance of either outsourcing your data center efforts or building your own reaches a critical level, firms on either side of the business-size spectrum are looking at finding the best value for money to store and manage their information. However, trying to find value but opting for a cheaper service seems to be an all-too-easy route for small and large businesses to take, causing extreme concern. Instead, with just a little research about what they should be looking for in a data center, firms can avoid the 'get what you pay for' providers and uncover real value in the market.

Which provider?
Before selecting a data center, firms need to decide which provider will deliver the services they need. However, with huge competition in the market, it's important to research differentials that separate the exceptional from the average. For example, if a firm is a Microsoft Certified Partner, it has demonstrated its expertise and skills in its chosen field - as well as passing a number of examinations set by the tech multinational. As a result, the firm would be fully qualified to deliver Microsoft products and services in the eyes of Microsoft.

Furthermore, if businesses hoping to outsource their technical infrastructures are worried about energy efficiency and the environment, it may be helpful to look for firms that are members of The Green Grid. A not-for-profit consortium, The Green Grid seeks to corroborate green ideas and services by these firms in order to improve resource efficiency. As a member of the group, a business can be sure that their chosen provider is a leader in providing resource efficient data centers.

Arrange a visit
After researching the credentials of a web hosting or data center provider, the next step would be to physically see the data centers they provide, firsthand. This way, prospective clients get an idea of the location and the physical security installed by the provider. While most data centers are located in city centers and business parks (meaning their owners can guarantee what is happening within the confines of the property), rural locations are also viable, as they can offer protection away from very busy, populated areas.

In terms of security, a business may be asked to register its information beforehand, indicating one of the many layers of security a data center may harbor. Other layers can include keycard access, biometric fingerprint readers and 24/7 surveillance teams monitoring the perimeter in order to prevent unauthorized access. Retinal scanning can also be another useful tool. Generally, military-level security personnel keep watch over the perimeter, while three-step authentication ensures only recognized staff can access technology. All of these measures combine to highlight the importance of a customer's data.

Overall, customers should be checking not only the actual services that a data center provider can offer but the actual provider itself. Awards, certifications, testimonials and badges can offer an authoritative edge in a saturated market, while visiting the data centers provides clients with the chance to visualize what they are paying for. Getting value for money is critical in the quest for a data center, but remember - it's an investment worth making, rather than one that should be done cheaply.

Dominic Monkhouse is the UK Managing Director at PEER 1 Hosting.