Tags

Can Your Data Center Predict the Future?

Who could have predicted that in the span of a few weeks, the COVID-19 pandemic would transform the planet into a stay at home workforce? As a result, data center remote management and monitoring capabilities have moved squarely to the forefront. Large enterprise and government organizations are scrambling to identify tools and processes that optimize all aspects of their IT infrastructure for the new remote workplace.

Who could have predicted that in the span of a few weeks, the COVID-19 pandemic would transform the planet into a stay at home workforce? As a result, data center remote management and monitoring capabilities have moved squarely to the forefront. Large enterprise and government organizations are scrambling to identify tools and processes that optimize all aspects of their IT infrastructure for the new remote workplace.

Managing power is one of the most critical tasks carried out continuously in any data center. The slightest changes can have significant consequences, affecting temperature levels and server performance. Power is also one of the most expensive costs in the data center making it one of the most scrutinized areas of focus as organizations evaluate their remote data center management processes. Without real-time visibility into power statistics, customers run the risk of miscalculating consumption, purchasing more power than they need, or more equipment than their data center space can support.

Only through digitization of IT and physical data can data centers begin to apply AI, machine learning and predictive analytics that are fueling the improvements in visibility required for accurately forecasting power utilization. 

Power Utilization Tied to Changing Needs

Data centers are dynamic IT environments that undergo constant changes based on the changing needs of its customers. Increasing server density affects power requirements that may result in changes in cooling demands. Compute and storage requirements change based on demand. Colocation data center racks are constantly being added or moved to accommodate new customer equipment and every change can have a complex waterfall effect.

This complexity makes it increasingly difficult to anticipate potential consequences of any change. What happens when a server is moved from one rack to another?  What if one part of an aisle experiences unexplained changes in temperature?  How do these changes affect my goal to maintain 80% power utilization?

Emergence of Software-Defined Service Delivery

Following several years of intense focus and execution, QTS has successfully accomplished its goal of digitizing its external and internal systems.  QTS’ API-driven Service Delivery Platform (SDP) is the industry’s first orchestration platform that empowers customers to interact with their data and QTS services by providing real-time visibility, access and dynamic control of critical metrics across hybrid IT environments from a single platform.

QTS’ 17,000 active SDP users have access to key information and control over critical infrastructure, data and processes. And as COVID-19 shifts nearly all of us into a stay at home mode, this flexibility to remotely provision services and capabilities is aligning with the requirements of an expanding remote workplace.

QTS’ Power Analytics is an SDP application that will apply predictive analytics, utilize statistical algorithms and AI-driven machine learning techniques to analyze real-time power data gathered over time to anticipate future outcomes. It detects when a customer is approaching amperage thresholds and automatically notifies them.

SDP’s predictive capabilities makes QTS one of the first providers to show a customers’ consumption versus its committed power. Customers learn in real-time if they are under or over consuming and are able to make changes to optimize performance and spend. In 2019, QTS reported that SDP Power Analytics generated 63,510 automated notifications of power overages.

By accessing accurate, real-time power draw by location, suite, cage, rack, circuit and even pole levels, SDP Power Analytics users have the insight to improve and validate actual-to-plan capacity, to optimize space and rack allocation, and to identify circuits and poles that may only need balancing.

Power Analytics - Key Capabilities

  • Understand total power consumption vs power contracted by site.
  • Receive alerts as enclosures and circuit pairs reach thresholds.
  • Identify under and oversubscribed circuits, racks and cabinets.
  • View power trending; identify fluctuations
  • Export power usage reports
  • US SDP at the data center to validate configuration changes

Real World Benefits

 

One of the world’s largest semiconductor companies chose to colocate in a QTS data center to deploy critical production environments in a high-density environment. Power Analytics is providing them near real-time power metrics allowing for greater operational efficiency and improved forecasting. It is enabling the company to retrieve and analyze power metrics related to their QTS deployment by location, by nested spaces, and power devices including PDUs, UPS’, and circuits down to the rack and pole level.

 

According to the company’s data center manager, “Our goal is to maintain a maximum 80% utilization of power across our environment. Before SDP I was getting a weekly report from the team but it did not break down utilization at the rack level. I could only know, collectively, what our utilization was but could not know what, if any, racks were being oversubscribed. It was difficult and time consuming to identify which servers were over or underutilized so we were forced to purchase additional power to ensure we would avoid any interruptions.”

With Power Analytics, the semiconductor giant is now analyzing the following:

  • Power consumption in its data center footprint versus its power subscription - useful for planning future expansions
  • Power consumption at the rack level versus capacity of the PDU – goal is to optimize under 80% of the manufacturer’s value.
  • Power consumption on each of the two PDUs on a rack – ensures that a single PDU is sufficient to handle the load in case a PDU fails
  • Be notified proactively via email if any 80% capacity threshold is threatened

 

“Today we use Power Analytics on a very regular basis to keep our environment normalized and optimized for the 80% utilization threshold, as well as for future expansion and planning.”

While no one could have predicted the extent to which the COVID-19 pandemic is impacting all of us, digitization of data and predictive analytics are now enabling data centers to be better prepared for future problems. This preparation helps to reduce costs and allows data centers to invest in other areas to serve customers more effectively. Most importantly, predictive analytics can deliver measurable benefits in the form of better power utilization, allowing data center customers to optimize their IT and focus on growing their business.

For more information on QTS’ SDP and Power Analytics, check out this short video.