What is stranded capacity?
Capacity is a term we often use to measure the available power, cooling, and space within a data centre. Stranded capacity is installed capacity (cooling, power, space) that cannot be used to support the critical load. When assessed individually each of these parameters (cooling, power, space) do little to depict the actual capacity of the data centre. As each parameter is interrelated, they must be assessed as a whole to ensure a balance is maintained between each. Stranded capacity occurs when an imbalance occurs between the 3 parameters that restrict the overall power efficiency of the data centre.
Power and cooling capacity become stranded when predicted IT power densities never materialize. For example, many sites that were designed for 100 W/sq. ft or more continue to operate at 50 W/sq. ft or less. In this case whilst there is limited physical space available, 50% of the rack power is still available resulting in stranded power as there is no space to install additional infrastructure at the current rack density.
As racks are not operating at full power density, the cooling capacity installed to support the unused 50% of the load is also stranded. Vice versa, available power can become stranded due to inability to install cooling infrastructure where it is needed. When it comes to stranded space capacity, this typically comes down to footprint of infrastructure, the larger the footprint the less W/sq. ft of available power.
The bottom line is that stranded capacity reduces the operating efficiency of the data centre and should be carefully monitored to ensure data centres are operating with an optimum power, cooling and space balance.
The Problem with Traditional Data Centre Capacity Planning
As previously highlighted, the key to avoiding stranded capacity is striking an efficient balance between the available space, power, and cooling capacities. The actual capacity of the site will be limited by the most restrictive parameter of the 3.
However, accurately planning data centre capacity is easier said than done, with the constant balancing act between power, space and cooling presenting a challenge in the design and modification of data centre facilities.
Traditionally data centre capacity planning relied on this formula:
Future Resources = Current Usage x (1 + Normal Growth + Planned Growth) + Headroom
Although the use of a standard formula would seem to simplify data centre capacity planning, it does quite the opposite as it fails to account for the variability of load on the servers. Granted this formula is fine for standard office buildings and other static loads but data centres require a more advanced method of planning. Critical load requirements in the data centre can be very unpredictable due to rapid changes in demand and as the demand for data increases, data centre infrastructure must evolve to facilitate higher power capacities and cooling requirements. If a static formula is relied upon for capacity planning when demand is so volatile, stranded capacity can almost be guaranteed as power requirements continue to grow and the balance between power, space and cooling become mismatched. It is important to note that data centre capacity planning is not a one-time task, but instead it is a continuous process of monitoring and modification to strike the optimal balance to support efficient data centre operations.
Avoiding stranded capacity
To avoid stranded capacity you must first identify the limiting factor and modify the capacity of the remaining 2 elements to rebalance each of the 3 defining parameters.
Space as limiting factor
When physical space within a data centre facility has been exhausted, modular e-houses are an efficient solution to facilitate continuous upscaling of power capacity. Additionally, the introduction of low footprint infrastructure within the data centre can help to optimise the white space and eliminate wasted square footage.
Power as limiting factor
Overbuilding of data centre capacity is a major issue and is largely due to the notion that data centres need to be armed with enough capacity to meet unforeseen demand. In reality, this can lead to excessive stranded capacity which can be very costly. It is important to right size your data centre to support optimal operating efficiency where cooling, power and space are in balance.
Cooling as limiting factor
If cooling capacity is not able to efficiently cool the power load, the data centre’s PUE will suffer; systems will become overheated and some of the available power will be wasted as heat output.
Underfloor cabling with limited space for cooling can contribute to stranded power capacity.
Power is critical to the reliability and availability of IT services. As a result, data centre owners are under mounting pressure to continuously monitor their infrastructure to plan for future capacity requirements and ensure system health to mitigate downtime. Data centres are very complex entities and therefore manual procedures and limiting calculations must not be relied upon to manage infrastructure requirements. It is vital that data centre owners have the tools required to better leverage their data to make informed capacity planning decisions that support data centre efficiency
Thanks to DCIM software, data centre capacity planning has been revolutionized enabling the automation of system monitoring and data analysis to facilitate accurate and fluid planning where capacity is optimized to the needs of IT and business services.
Alongside DCIM software, successful and efficient capacity planning can be done in the following steps:
- Outline all of the required components for new equipment to be implemented and installed in the facility.
- Measure the current usage level of the required component. Are they fully utilised or do they have additional capacity?
- Create a list of capacity requirements
- Build a plan, outlining how you will provision for the new equipment
- Issue work orders to physically provision the new equipment
- Audit the provisions and update the DCIM database to show the level of capacity utilization.
