Presented by Solidigm
As AI adoption grows, data centers face a significant bottleneck in storage space — and traditional hard drives are at the center of this problem. Data that previously lay dormant as cold archives is now being used more frequently to build more accurate models and provide better inference results. This shift from cold data to warm data requires low-throughput, high-throughput storage that can handle parallel computations. Hard drives will remain the backbone of low-cost cold storage, but without rethinking their role, the high-capacity storage layer may become the weakest link in the AI factory.
"Modern AI workloads, combined with data center constraints, have created new challenges for SSDs," says Jeff Jankowitz, vice president of research at IDC. "While hard drive vendors address the growth of data storage by offering larger drives, this often comes at the cost of slower performance. As a result, the concept of “nearby SSDs” has become an increasingly important topic of discussion within the industry."
Today, AI operators need to maximize GPU usage, manage network-attached storage efficiently, and scale compute – all while cutting costs on increasingly scarce power and space. In an environment where every watt and every square inch matters, success requires more than just a technology update, says Roger Correll, senior director of AI and marketing leadership at Solidigm. It calls for a deeper reorganization.
“It speaks to the fundamental shift in the value of data for AI,” Correll says. “This is where high-capacity SSDs come into play. Along with capacity, they provide performance and efficiency – allowing exabyte-scale storage lines to keep up with the constant pace of data set size. All of that consumes power and space, so we need to do it as efficiently as possible to enable more GPU range in this constrained environment.”
Not only do high-capacity SSDs replace hard drives, they remove one of the biggest bottlenecks in the AI factory. Delivering huge gains in performance, efficiency and density, SSDs free up the power and space needed to push the GPU’s range even further. It’s not so much a storage upgrade as it is a tectonic shift in how data infrastructure is designed for the age of AI.
HDDs vs. SDDs: More than just a hardware update
Hard drives have impressive mechanical designs, but they are made up of many moving parts that use extensively more power, take up more space, and fail at a higher rate than solid-state drives. Reliance on spinning platters and mechanical read/write heads inherently limits input/output operations per second (IOPS), creating bottlenecks for AI workloads that require low latency, high concurrency, and sustained throughput.
Hard drives also suffer from latency-sensitive tasks, where the physical act of searching for data introduces mechanical delays that are inappropriate for real-time AI inference and training. Moreover, their power and cooling requirements increase dramatically under frequent and intensive data access, reducing efficiency as data volume increases and temperature increases.
In contrast, VAST’s SSD-based storage solution reduces energy use by about $1 million per year, and in an AI environment where every watt counts, this is a huge advantage for SSDs. To illustrate this, Solidigm and VAST Data completed a study examining the economics of storing data at the exabyte-quadrillion byte, or billion gigabyte, scale, analyzing storage power consumption versus hard drives over a 10-year period.
As a starting point of reference, you’ll need four 30TB hard drives to equal the capacity of one 122TB Solidigm SSD. After taking into account VAST’s proprietary data reduction techniques made possible by the superior performance of SSDs, the exabyte solution includes 3,738 Solidigm SSDs versus more than 40,000 high-capacity SSDs. The study found that VAST’s SSD-based solution consumed 77% less storage power.
Reduce data center footprint
"We ship 122TB drives to some of the world’s top AI OEMs and leading cloud providers," says Coryell. "When comparing a 122TB SSD with a hybrid HDD + TLC SSD configuration, they get a nine-to-one savings in data center space. And yes, it’s important in these massive data centers that are building their own nuclear reactors and signing huge power purchase agreements with renewable energy providers, but it’s increasingly important as you get into regional data centers, on-premises data centers, and all the way to your edge deployments where space can come at a premium."
Nine-to-one savings go beyond space and power – it allows organizations to fit infrastructure into previously unavailable spaces, scale up the GPU, or create smaller footprints.
"If you were given X amount of land and Y amount of energy, you would use them. You are artificial intelligence" “Since every watt and square inch counts, why not use them in the most efficient way? Get the most efficient storage on the planet and enable more GPU range within that envelope that should fit you,” explains Corel. “On an ongoing basis, it will save you operational cost as well. You have 90 percent fewer storage slots to maintain, and the cost associated with that disappears.”"
Another element that is often overlooked is that the (much) larger physical footprint of data stored on mechanical hard drives results in a larger building material footprint. Collectively, concrete and steel production account for more than 15% of global greenhouse gas emissions. By reducing the physical footprint of storage, high-capacity SSDs can help reduce embodied emissions from concrete and steel by more than 80% compared to HDDs. At the final stage of the sustainment life cycle, the end of the engine’s life, there will be 90% less motivation to dispose of it. .
Reshaping cold storage and archives strategies
Moving to SDD isn’t just a storage upgrade; It’s a fundamental reorganization of data infrastructure strategy in the age of AI, and it’s gathering pace.
"Big scalers are looking to get the most out of their existing infrastructure, doing kinks, if you will, with hard drives like overprovisioning them to nearly 90% to try to extract as many IOPS per terabyte as possible, but they’re starting to show," says Coryell. "Once they shift to modern, high-capacity storage infrastructure, the industry as a whole will be on this path. Additionally, we’re starting to see these lessons learned about the value of modern storage in AI applied to other sectors as well, such as big data analytics, high-performance computing, and many others."
Although all-flash solutions are almost universally adopted, there will always be a place for hard drives, he adds. Hard drives will continue to be used for uses such as archiving, cold storage, and scenarios where the pure cost per gigabyte outweighs the need for real-time access. But as the token economy heats up and companies realize the value of data monetization, the hot and cold data segments will continue to grow.
Solving future energy challenges
Now in its fourth generation, with over 122 cumulative exabytes shipped to date, Solidigm’s QLC (Quad-Level Cell) technology has led the industry in balancing higher drive capacities with cost efficiency.
"We don’t think of storage as just storing bits and bytes. We’re thinking about how to develop these amazing drives that can deliver solution-level benefits," says Coryell. "The shining star of this is the recently launched E1.S, designed specifically for dense, efficient storage in direct-attached storage configurations for next-generation fanless GPU servers."
The Solidigm D7-PS1010 E1.S is a remarkable achievement, the industry’s first SSD with single-side direct-to-chip liquid cooling technology. Solidigm worked with NVIDIA to address the dual challenges of heat management and cost efficiency, while delivering the high performance required for demanding AI workloads.
"We are rapidly moving to an environment where all critical IT components will be cooled directly to the chip on the direct connection side," He says. "I think the market needs to look at their approach to cooling, because the energy constraints and energy challenges will not abate in my lifetime, at least. They need to apply the new cloud mindset to how they design the most efficient infrastructure."
Increasingly complex inference puts pressure on the memory wall, making storage architecture a front-line design challenge, not an afterthought. High-capacity SSDs, combined with liquid cooling and efficient design, are emerging as the only path to meeting rising AI demands. The task now is to build an infrastructure not only for efficiency, but also for storage that can scale efficiently as data grows. Organizations that reorganize storage now will be the ones able to scale AI tomorrow.
Sponsored articles are content produced by a company that pays for the post or has a working relationship with VentureBeat, and are always clearly labeled. For more information, call [email protected].
[og_img]
Source link