Saving energy and reducing the CO2 footprint in data centers is no longer optional

Imagine the parking lot in front of your office. All the cars are idling and have the air conditioning switched on, day and night, throughout the year. They are just waiting around for a passenger. Many of the cars are unattended. No one knows who owns them, not least the car park manager. Sounds silly, doesn’t it? This is exactly what is happening in many data centers and server rooms today.

In light of the Paris Agreement and the United Nations’ 2030 Agenda for Sustainable Development, which provides the foundation for sustainable, low-carbon, and resilient development under a changing climate, companies must report on their CO2 footprint annually.

For this reason, newly appointed sustainability representatives are now scratching their heads on many challenges, such as how to measure, report and improve their CO2 footprint.

Today, we are also in the middle of an energy crisis caused by the war in Ukraine, geopolitics around gas supply, and local grid capacity issues caused by the ongoing electrification of our society. The energy market is very uncertain, and prices are skyrocketing.

As a result, energy savings and related CO2 footprint reductions are no longer without obligation. Several governments in the EU have already announced and enforced new laws for companies to deliver energy consumption reports, provide details on power-saving settings in equipment and enforce energy-saving goals.

How and where can we save energy in IT?

It is estimated that 30% of IT equipment in data centers is running with low to 0% utilization. These “ghost” devices keep their own operating systems alive, ready for requests that may not be coming.

Ghosts are machines that do not seem to have an owner or are not actively monitored. Depending on configuration and power-save settings, these machines consume at least half of the power while idling as they do when fully utilized. Further, they generate heat that must be cooled down by the heating, ventilation and air conditioning (HVAC) systems, increasing the total power consumption even more.

Low-utilized hardware is often a result of companies moving their business-critical applications to the cloud but “forgetting” to decommission the old legacy servers or leaving some servers switched on because of some small, locally-used application that could not be moved. Legacy servers and connected network appliances needed for only a few small applications are very energy inefficient. Consolidation of small “left-over” applications in fewer servers can reduce energy consumption significantly.

Power Usage Effectiveness (PUE) indicates the efficiency of a data center. This KPI is normally owned by the DC Power & Infrastructure manager, who manages the facility. PUE is calculated by dividing total facility power consumption by IT equipment power consumption. The lower this figure, the better a data center performs. However, a major flaw exists in using PUE to indicate efficiency across the IT technology stack. The data center facility manager does not have any influence on the energy consumption of IT equipment as hardware sourcing, power-save configuration, utilization and workload management usually are managed elsewhere in the organization.

When IT equipment power consumption decreases because the teams responsible for IT equipment have optimized power-save settings or workloads, the PUE gets worse as overhead in the data center will not change much. When PUE is used as a KPI in a top-level dashboard to indicate IT use efficiency, management receives a wrong perception of energy efficiency in their organization.

To reduce energy consumption, costs and CO2 footprint, firms need to monitor the current consumption related to “useful” compute and storage utilization in the most granular way.

A holistic view is needed that covers technology stack levels, business processes and management domains, as follows:

Building and infrastructure level: Power and temperature metering and data collection using IoT sensors that can measure power consumption and power quality up to the device socket and detect phase load imbalances, power factor efficiency, idle IT equipment, etc. Other sensors pick up temperature and humidity inside and the weather outside the building. Power and infrastructure staff will get insight into the efficiency of the building and its overhead facilities and take preventative action based on trends.

Per device and/or OS level: Modern IT equipment, such as server hosts, have built-in sensors that can measure parameters such as power consumption, fan speeds, and inlet and outlet temperatures. Server operating systems are also able to monitor CPU, RAM, storage and network utilization. This will provide the service operations teams with insight into how the hardware performs against utilization.

On (virtual) workload level: Hypervisor management software can provide CPU/RAM/traffic utilization and even (calculated) power consumption readings per host, virtual machine, or workload in hypervisor resource groups. With this information, it is possible to move workloads and virtual machines around and find utilization sweet spots within a group of hosts by taking into consideration performance and capacity buffers, virtual machine spin-up times and day/night/weekend utilization effects. Trending and prediction analysis of utilization data helps minimize excessive host capacity by putting hosts in power-save mode or switching off to save energy when capacity is not needed.

IT asset management: Not every IT asset, including virtual machines, installed in an organization is accounted for and managed from a monitoring and life-cycle point of view. In many cases, IT assets are managed in spreadsheets by local IT organizations or equipment service owners without providing a global view. There are many ways to scan your network to find devices with an IP address and keep an online database or CMDB. However, you will never have a full picture of devices connected as misconfiguration, un-responding (OT) equipment or firewalls may block scans, etc. In those cases, physical site surveys are needed to check what is really in the cabinets to detect ghost devices. These unknown and/or illegal devices on your network not only consume unnecessary energy but may also pose a serious security risk, making proper IT asset management a key function.

Going to the cloud

Companies have moved many applications to the cloud where insight into power consumption and related CO2 footprint became less important because the responsibility to manage energy consumption in devices moved to the big hyperscalers. However, many industries still need to report on their CO2 footprint like, “How much CO2 is one financial transaction” or “What is the CO2 footprint of storing one Gigabyte of data for a year” or “How much CO2 is needed to produce a bike.” The hyperscalers provide dashboards for carbon footprint but cannot always relate to a business function, process or production batch.

Consolidate and analyze data

Administrative systems managing your business processes, such as ERP systems controlling production, ordering, sourcing, logistics and service and incident management systems contain valuable data. Administrative data correlated with technical data from IIoT metering, monitoring and device management systems gives valuable insight in the efficiency of an end-to-end business or production process, including external factors such as incidents, weather, seasonal effects and economic changes, so it can be optimized.

Talking to your process managers and using AI and ML techniques help to understand what the “perfect” process should look like from an efficiency and power consumption point of view by looking at every element, circumstance and configuration in any process.

How can Orange help?

We have learned from our own experience that measuring and analyzing energy consumption, environmental parameters and business process KPIs are mandatory to address efficiency and optimization actions that can lead to CO2 footprint and cost reductions.

Successful energy efficiency management demands an accurate understanding of business processes and the applications and infrastructure used in your organization. This can be very comprehensive and hard to manage. Therefore, choosing your battles and carefully defining your scope and objectives is mandatory.

The chosen approach is based on a holistic view and can contain many pre-defined building blocks that we have at Orange Business, including:

  • Digital Green Maturity Assessments to support the definition of scope and objectives
  • Setup monitoring and metering
  • Identification of administrative data sources
  • IT asset management and physical inventory site surveys
  • Utilization of service management insights
  • Data collection and interfacing
  • Data consolidation and modeling
  • Data science and dashboards
  • AI and ML consultancy and development
  • Quality feedback into your processes

Contact our Digital Consultants who can help you find the right strategy among the blend of solutions available.

Marcel van Wort

Marcel is the Managing Consultant specialized in Information Security and Green IT at Orange Business with more than three decades of experience in ICT. He is an active member of the Orange Green Act program committed to achieve Green IT, sustainability and CSR goals at Orange Business and help our customers with their digital sustainability transition. Marcel likes mountain biking and is on a mission to develop ways to use CO2-neutral bio-fuels in motorsports while racing the Dakar Rally 2023.