Grid reliability: powering change in the data centre
Thu 24 Jun 2021
Data centres consume enormous amounts of power –far more than their fair share of energy from local electric grids.
Data centre owners and operators are looking for innovative solutions to ensure grid reliability, which helps them provide stable and resilient services to meet high customer expectations.
A study from Frost & Sullivan found that 81% of data centre operators are looking for a higher standard of power grid reliability than what they currently experience. However, these companies are not depending on local infrastructure to improve reliability – rather, they see investing in their own solutions as the best way to ensure their data centres have the power they require. In fact, 52% of data centres already spend $50,000 or more per annum to improve power reliability.
The Current Situation
Industry experts estimate that globally, data centres consume around 205 TWH per year – roughly 1% of available electricity. And over the past year, pandemic-related societal changes have drastically increased demand for data centre services. For example, during the initial lockdown in March-April 2020, internet traffic increased by 25-30%, 10x the normal growth rate. Demand for cloud-based services increased as well, as did the demand for big data, AI and ML-driven solutions, and IoT enablement, particularly in healthcare.
The assumption then would be that as demand for data centre services increases, the stress on the power grid should increase as well. However, this number has been fairly stable over the last decade, rising only 6% since 2010.
This may be attributed to a variety of factors, including:
- More energy-efficient IT hardware
- Adoption of server virtualization, improving usage
- Migration to cloud and hyperscale
- Innovative cooling techniques
205 TWH is still an enormous amount of power to demand from a local grid. And all grids are not created equal: in some areas, reliability is affected by an ageing infrastructure and in others, a lack of developed infrastructure.
For example, in the U.S., the original power grid pathways were created in the early 1900s, and utility companies are still running systems that were built 50-70 years ago. An average of 500,000 people are affected by power outages every day, highlighting the need for data centre providers to supplement the local grid with power backups and energy-efficient, innovative solutions.
Recently, the Irish government reviewed the agreements that their national grid management company had made with data centre providers requesting new connections to the power grid. To protect the grid in already-stressed urban areas, the Commission for the Regulation of Utilities (CRU) directed Eirgrid to prioritise connections in rural areas.
Regions in which data centre providers build facilities are separated into tiers, generally Tier I, Tier II and Tier III. Tier I and II have been fairly well-developed, and data centre businesses are looking to move to the developing regions in Tier III. For example, Alibaba Cloud recently announced the $1B Project AsiaForward, building data centres and providing technological skill-building programmes to regions in South East Asia.
However, the Tier III markets tend to be in areas where the infrastructure is less developed, and grid reliability is low. For instance, the African data centre market is expected to exceed $3 billion by 2025, even though many areas are not yet cloud-ready.
The Need for Grid Reliability
People depend on online services, provided through businesses that depend on data centres. And data centres depend on the grid, to provide the power needed to reduce downtime for their customers.
As data centre technology has improved, customer expectations are getting higher. Not long ago, three 9s (or 99.9% uptime) was the industry standard. However, that is no longer a competitive estimate: with SLA’s commonly promising five 9s (99.999%) or even up to seven 9s (99.99999%) of uptime.
The Future of Data Centre Power
To remain competitive and manage customer expectations in the face of ageing or underdeveloped infrastructure, companies are investing in their own solutions. In addition to improving efficiency, data centres are looking to renewable energy sources and innovative cooling solutions to lessen dependence on the grid, as well as improved backups for increased reliability.
Advances have been made in renewable energy, increasing availability and reducing costs. Many data centre companies are looking to deliver on-site renewable energy sources, to improve reliability and reduce stress on the grid. Others are building strategic alliances with renewable energy providers to utilize solar, wind, and other renewables at their sites; or by entering into more complicated virtual PPAs (Power Purchase Agreements).
Keeping servers cool uses an inordinate amount of a data centre’s total power. This has been combatted by placing facilities in areas that are naturally cooler – in Iceland, underground, etc. New advancements in cooling are being moved from the testing phase and into practical applications as well: for instance, the underwater server farm that Microsoft tested will be a product offering in China later this year.
Immersion cooling is being used in more live server environments, as is liquid cooling. These technologies allow companies to create new data centre facilities in regions where they are needed, to improve local access – while still helping to control costs and reduce demand on the power grid.
Finally, a large part of grid reliability involves coping with the occasional interruption in power – which is why data centres are built with redundancy protections, commonly diesel-powered backup generators. However, to meet new standards of lessening environmental impact, many companies are moving away from diesel-powered backup generators, and instead are exploring hydrogen fuel cells and even lithium batteries as practical alternatives.
Grid reliability is a top concern for data centre providers, and for the companies and people that rely on data centre services. As demand continues to grow worldwide, and companies enter different areas of the world, alternatives to placing additional stress on local power grids will be of more importance to ensure grid reliability.