Datacentre industry stakeholders and regulators are expected to expand their energy efficient datacentre criteria to include assessments of IT equipment utilisation and work delivered per unit of energy consumed.
While the industry has traditionally used Power Usage Effectiveness (PUE) as the proxy metric for datacentre efficiency and sustainability, it does not address the efficiency of the IT equipment.
And there is a realisation among operators and industry stakeholders that truly efficient datacentres must maximise the utilisation of the IT equipment work capacity and minimise the power demand to the greatest possible extent.
This shift in focus can be seen in the comments filed to the draft of the EU Energy Efficiency Directive (EED) Delegated Regulation (November 2023). Several commenters encouraged the EU Commission to require reporting of the average utilisation of the installed server and storage equipment to understand how effectively their available work capacity is being used. They posited that excluding this key performance indicator (KPI) from the EED data collection process resulted in an incomplete picture of a datacentre’s performance and efficiency.
From an economic standpoint, datacentre operators benefit from maximising equipment utilisation and reducing energy use. Higher utilisation reduces the IT equipment needed to deliver a defined workload, reducing required space, energy consumption, and capital expenditures.
Operators can mitigate the impact of increased IT workload requirements by refreshing and consolidating existing workloads onto more efficient IT equipment, releasing power, cooling, and space resources to accommodate new workloads.
Uptime Institute data from its IT power and efficiency survey 2023 indicates that operators track, collect, and analyse the utilisation profiles of their server infrastructure and set performance goals to achieve high levels of server utilisation.
IT operators are hesitant to reveal the average utilisation of their datacentre IT equipment. In some cases, equipment utilisation could be higher due to the misplaced view that links overcapacity to increased operational resiliency and reliability. In other cases, it is due to regulatory or business resiliency and reliability requirements, such as the need for financial institutions to run twin datacentres for regulatory purposes or to manage workloads with short-term bursts of capacity demand.
In either case, IT operators have access to tools and business processes that will enable them to improve and maximise the utilisation of their IT equipment while maintaining or improving their operational resiliency and reliability.
Applying these tools while setting utilisation goals for the server infrastructure can drive surprising improvements in server efficiency and performance. Data analysis has shown that server work per watt improves by 2.5 x to 3.5 x when average server utilisation is improved from 12.5% to 50%, with utilised workload capacity increasing by 4 x.
Increasing average server utilisation requires organisational commitment. Only 46% of operators in the IT power and efficiency survey 2023 have set goals to improve their average server utilisation.
Figure 1: Less than half of IT operators have utilisation improvement goals
Almost half of those operators that set goals create a stretch goal of 65% utilisation or greater, and two-thirds strive to achieve average utilisation of their server infrastructure of more than 50%. If an operator starts at 20% utilisation, increasing to over 50% can reduce the physical server count by half or more.
Figure 2: Operators set aggressive average server utilisation goals
The results accomplished by setting aggressive goals are impressive. For example, 67% of the operators that set utilisation goals achieve average utilisations above 50%, and 92% are above 25%. Knowledgeable datacentre industry experts report that industry average server utilisation ranges from 20% to 30%.
These operators, who have set utilisation objectives and executed improvement plans, are operating well above the industry average with attendant capital, expense, and space savings. They are also operating a demonstrably more efficient and sustainable datacentre.
Figure 3: The average utilisation of the server fleet
Unfortunately, the survey data shows that 67% of operators do not track and calculate individual server utilisation levels (Figure 4). The lack of this information makes it difficult to set goals to drive improvements. This deficiency exists across the full range of datacentre sizes, with only 24% of operators of large datacentres (more than 10 MW) tracking utilisation at the server level.
Figure 4: The percent of operators tracking server utilisation by power capacity
Given the expected push toward establishing requirements for the public reporting of IT equipment utilisation levels, operators are advised to begin collecting and analysing utilisation data. Collecting the baseline data will enable an operator to set a server utilisation goal, identify, classify (easy to hard), and prioritise consolidation projects, and establish a multi-year consolidation plan to achieve the goal.
For most operators, basing the plan on a ‘refresh and consolidate’ strategy will facilitate the attainment of a server utilisation goal.
Using the refresh process, the IT team can consolidate workload from two to three (or more) existing servers to one new server as part of the migration process. By consolidating on refresh, the savings in new equipment, energy, and operating space can fund the IT work needed to create or upgrade the virtual machines and containers, optimise workload placement, and validate that application service levels are achieved in the new, denser server-level application environment. The improved operational environment, environmental performance, and reduced costs will please everyone from the CEO to the chief financial officer to the IT operator to the sustainability manager.