How to Find Stranded Capacity in Your Data Center
Data center capacity planning is one of the biggest challenges for today’s data center professionals. According to a recent survey by Sunbird Software, 72% of respondents said that capacity planning was one of their top objectives.
Proper capacity planning results in the right-sized data centers, efficient utilization of resources, and reduced costs, but it is easier said than done. The reality is that many data center professionals struggle to accurately plan capacity and instead have rampant stranded capacity.
What Is Stranded Capacity?
Stranded capacity refers to data center resources that are not known to be available for use. Most of the time, you don’t even know if you have stranded capacity and how much there is.
Power, space, cooling, and data/power ports are the main capacity parameters in the data center. When an imbalance of resources occurs, stranded capacity is present. For an accurate picture of capacity, all resources must be viewed as interrelated and interdependent to understand the current balance.
Why Is Stranded Capacity Important?
Stranded capacity is a waste of resources and money. It reduces the efficiency of your data center, leading to higher operating costs and unnecessary new buildouts.
It costs about $15,000 to deploy a server-ready cabinet. If you can use stranded capacity for new deployments instead of provisioning new resources, you can save a lot of money. Imagine the savings you can utilize by leveraging existing resources and deferring the need for new cabinets or even a new data center.
Try our What Does It Cost to Power a Rack? Calculator to see how much money you can save by recovering stranded power capacity instead of building out new racks.
Similarly, cooling a data center is expensive, and getting the most out of your existing cooling capacity can save a lot of money.
What Causes Stranded Capacity?
Common practices and mistakes in the data center that can lead to stranded capacity include:
- Overprovisioning of data center capacity. When organizations design their data centers, they often include a large excessive amount of capacity beyond what they require. It’s wise to be prepared for future growth and unforeseen circumstances, but too much overprovisioning is wasteful.
- Traditional planning methods. When budgeting power and determining how many servers can be deployed in a cabinet, data center managers often derate the nameplate value of a server to around 60-70%. The resulting budget numbers are just assumptions and can be highly inaccurate.
- Lack of tools and information. Planning capacity based on assumptions and manual calculations rather than reliable business intelligence and analytics leaves too much room for error that strands capacity. Legacy tools like Excel and Visio leave you susceptible to human error, are time-consuming, and suffer from poor version control.
- Ghost servers. Up to 30% of servers in the data center may be ghost servers (or zombie servers) that are physically running but do not perform any useful function. Ghost servers consume space, power, cooling, and cabling resources without providing any benefit.
- High-density deployments. High-density workloads involve a few power-hungry devices in a cabinet. There may be available space in the cabinet, but it cannot be used because deploying any additional devices would generate more heat than traditional cooling methods can accommodate.
- Inefficient airflow. If there are misplaced perforated tiles, unsealed cable openings, gaps in and around cabinets, cold air can escape. This strand's cooling capacity is because the cold air is wasted and no longer available to cool equipment. Similarly, not using a hot aisle/cold aisle containment strategy allows cold air to mix with hot air which reduces your cooling capacity.
- Overcooling. The average data center has 3.9 times more cooling capacity than the IT load requires. Overcooling wastes energy and money and results in unnecessary purchases of more cooling units.
- High relative humidity. In humid data centers, condensation can form on cooling unit coils. This can give off heat that uses a percentage of the cooling capacity.
7 Best Practices to Find Stranded Capacity
- Use modern data center management tools. You cannot manage what you cannot measure. Instrument your data center with intelligent rack PDUs and other metered power infrastructure and environmental sensors (i.e., temperature, humidity, pressure). Then, use data center management software to automatically collect, store, and trend the data. Modern DCIM software will transform that data into actionable insight such as how much-stranded capacity you have and where in the form of zero-configuration charts, reports, and visual analytics.
- Automate device power budgeting. Leverage DCIM with an Auto Power Budget feature that automatically calculates an accurate power budget number for each device instance based on the actual measured load of that device in your environment running your applications. You just have to set your policy once and the machine learning algorithm updates your power budget numbers each week. With this feature, Comcast gained 40% more utilization out of their power and facilities resources and eBay deploys projects with 33% fewer cabinets.
- Find ghost servers. Decommissioning ghost servers frees up the power, space, and cooling capacity that is being wasted on idle devices. With DCIM software, you can easily run a report to identify your potential ghost servers. You can then investigate and confirm that they can be decommissioned.
- Perform what-if analysis. What-if analysis helps you determine, on a per-project basis, the impact of additions and decommissions in the data center on your space and power capacity. With this information, you can know if you can use existing resources for future projects to delay purchasing more capacity.
- Intelligently search for capacity. With DCIM software’s intelligent capacity search feature, you can find where you have the capacity for new equipment within seconds. Simply select the model you’re deploying, and the software shows you every cabinet location that has available capacity for that specific piece of equipment.
- Ensure optimal airflow. Making sure that the airflow is as efficient as possible will reduce the amount of stranded cooling capacity in the data center. Follow data center airflow management best practices such as implementing hot aisle/cold aisle containment and installing blanking panels in unused spaces.
- Increase temperatures. Cooling unit manufacturers commonly rate their equipment based on a return air temperature of 75°F. However, data center managers commonly overcool equipment which does not allow the rated cooling capacity to be fully utilized. Raising temperature set points while monitoring that equipment is not exceeding manufacturer or industry guidelines saves money and reduces stranded cooling capacity.
Bringing It All Together
Stranded capacity wastes resources and must be minimized. Tackling the challenges of stranded capacity head-on can result in more efficient operations and significant cost savings.
With a modern data center capacity management solution, you can make more informed decisions that improve capacity planning and increase the utilization of existing resources.
Want to see how Sunbird’s second-generation DCIM software can help you recover stranded capacity? Get your free test drive today.