The key to unlocking Singapore’s sustainable AI ambitions? All-flash data centres
By Pure Storage
As Singapore seeks to position itself on the global AI stage, what does it take for public agencies to harness the benefits of AI while going green? Pure Storage makes the case for all-flash storage as the key to making AI systems more efficient.
Can Singapore achieve both its AI and sustainability goals? Pure Storage makes the case for all-flash storage as the key to making AI systems more efficient. Image: Canva.
Last year, Singapore unveiled plans to both green its data centres and accelerate artificial intelligence (AI) use in the country through new funding and a refreshed national AI strategy.
But according to a new study, AI servers could be consuming as much electricity as what countries like Sweden, Argentina, and the Netherlands use in a year. Can Singapore really achieve both its AI and sustainability goals at the same time?
“The demand for smarter infrastructure has never been more pressing,” says Chua Hock Leng, Area Vice President, ASEAN and Greater China from Pure Storage. Energy consumption is just one of the many AI challenges, he says, as public agencies often struggle with modernising legacy systems to support large datasets needed to also maximise the benefits of AI.
GovInsider sits down with Chua to understand how data storage can resolve the common AI challenges faced by public agencies, and how governments can choose the right storage solutions for their AI needs.
Flash storage can help maximise process and energy efficiencies
The type of data storage matters as data is key to maximising the benefits of AI. AI systems are trained on vast amounts of data to generate actionable insights that help support critical business decision-making, Chua says.
Data capacity continues to grow at around 30 to 40 percent yearly for most enterprises, Chua explains, and it may not be practical – in terms of floor space and power budgets – to continue adding hardware to their traditional storage infrastructure to accommodate more data needs.
Currently, 73 percent of American IT buyers find themselves not completely prepared for the energy requirements of AI, according to a recent survey conducted by Pure Storage.
This is why IT decision-makers will increasingly gravitate towards more efficient IT infrastructure, such as all-flash systems, that require less hardware and use less energy, he says.
“All-flash storage can deliver storage density that is two to three times better than competitors that are using commodity off-the-shelf solid-state disks (SSDs), while consuming up to more than 50 per cent fewer watts per terabyte than competitors. This can help enterprises reduce their total power usage by between five and 10 times,” he says.
“While SSDs technically use flash, they do not optimise the flash as much as Pure Storage’s DirectFlash, which enables the array to communicate directly with the raw flash,” explains Chua, on the difference between SSDs and Pure Storage’s all-flash storage solutions.
Chua likened SSD to a legacy hard drive: Like one contiguous set of identity blocks. In the case of DirectFlash, it uses a different approach that communicates with flash memory directly, maximising the capabilities of flash and providing better performance, power utilisation, and efficiency.
South Korea’s Chungbuk Technopark, for example, migrated to an all-flash storage solution to support its own AI initiatives as well as that of its tenants. This has led to a two-fold increase in speed when it comes to processing its stored data for AI workflows.
Meta has also tapped on all-flash storage for its AI Research SuperCluster, one of the world’s fastest AI supercomputers. Using all-flash storage, Meta stored and processed petabytes of data with low energy usage, allowing the company to invest more energy into its GPU, improving overall performance and speeding up AI models.
Petabytes is the unit of measurement for digital data, in which one petabyte is equivalent to 1,000 terabytes.
At the same time, all-flash storage also requires less maintenance than traditional disk storage, which has helped to reduce overall operational costs.
Governments need to break down legacies to tap into AI’s full potential
“AI implementation requires a completely new infrastructure to tackle the complexities of legacy solutions that prevent governments from tapping into the full potential of AI,” says Chua.
Chua shares that a single, scalable storage platform, such as that offered by Pure Storage, can help public agencies accelerate their paths to modern analytics and AI workflows. “Considering that AI will continue to accelerate in adoption, government IT teams will require an efficient, reliable, and high-performance infrastructure partner to ensure effective deployment,” he says.
As cloud or legacy infrastructure is often used in early AI development, public agencies need to consider the following factors to optimise their infrastructure for AI.
Firstly, public agencies need to consider the scalability and flexibility of data storage solutions for future growth. Current storage solutions should not only address current dataset sizes, but be able to accommodate large and expanding datasets for potential growth over time.
Secondly, democratising data access through a simplified and user-friendly platform is key for public agencies to fully benefit from data-driven insights.
Thirdly, as guardians of critical information and infrastructure, public agencies need to prioritise security, control and governance by ensuring that their data storage solutions offer protection against potential vulnerabilities and unauthorised access to sensitive data in AI applications.
“A solid data storage foundation that supports responsible and ethical AI implementation will foster trust and confidence in the use of AI applications across the public actor,” Chua says.