Artificial intelligence (AI) would not be possible without the technological advancement of data storage. But, now, AI is forcing us to question whether our current data storage capabilities are enough to handle its ever-expanding possibilities.
Huawei and the industry need to grasp the changes that AI brings to data storage, and it needs to do so very quickly, according to Peter Zhou, vice president of Huawei and the president of the data storage product line.
“When digitization started, data became information – as in data was being recorded, digitalized,” Zhou says. “People started to use big data to dig out the knowledge from the data. And then AI comes, which is where I think data will become ‘intelligence’.”
Huawei is a leading global provider of data storage products and services, and it emphasizes the importance of on-demand and ubiquitous storage, especially with the changing needs of data storage. Businesses now expect extensive convergence, adaptability, and a future-proof design for their storage architecture. Not to mention solutions that fit with their sustainability commitments.
To this end, Huawei provides a wide range of data storage products, with its New-Gen OceanStor Dorado Converged All-Flash Storage, OceanStor Pacific Scale-Out Storage, and OceanProtect Backup Storage. The tech company also offers a Data Management Engine (DME), and DCS AI Solution.
What’s more, Huawei also looks at what it calls ‘scenario-specific solutions’, which are essentially about designing and building tools that can be tailored to its customers' needs. A good example would be its container storage service for new application deployment, or its ransomware protection storage solution.
The value of data
For its data storage business, Huawei has over 26,000 customers from more than 150 countries and regions – protecting and using some 30,000 PB of data. These are customers that include healthcare providers, larger manufacturing firms, and even governments.
Huawei says its customers are suggesting that their data is increasing every year because of AI. These businesses have to store their data for much longer for training purposes. So the data itself - and its value - is increasing. As such, the demand for highly efficient data storage solutions is also rising.
Not so long ago, a petabyte of data would roughly take up a whole rack within a data center, but we are seeing multiple volumes of petabytes within a similar space. And yet still the demand is increasing with AI.
“The importance is getting higher and higher,” Zhou says. “By now everyone understands each country, each person, really cares about their data – data becomes essential. We believe that data storage will be more and more important. That’s the vision, so why not put more money into it to meet the requirements of the future.”
Here, Huawei is doing what Zhou says, with an extensive program of research and development for data storage. Huawei data storage has 12 centers for R&D and 8,000-plus R&D engineers. And that also includes 3,200-plus patents around the world.
“We need to keep developing the new generation of data storage,” Zhou says. “If you keep using the conventional data storage, the system still works, but with low efficiency. And that also means no commonization where only richer guys can do AI - which is not good.”
The warm data challenge
Typically, we see data in three ways: hot data, which is constantly in use. Warm data, which is stored but regularly retrieved and used. And finally, cold data, which is what we archive, often for many years.
“I personally believe, cold data will get smaller and smaller,” Zhou explains. “Warm data will get bigger and bigger. The reason is that all data is an asset, or knowledge, or intelligence. When AI comes, it will need to revisit not just current data, but also historic data. Which means the cold data, that portion will get smaller as the value gets higher and higher. The warm data portion will get bigger and bigger. And then the problem will be how to store the warm data. If warm data gets bigger, can solid-state drives (SSD) accommodate it? I personally don’t think it’s very suitable because of the price.”
High-capacity hard disk drives (HDD) are what we use to accommodate warm data, but the problem here is that HDD isn’t good for energy-saving. As Zhou puts it, HDD “keeps running”, and it has to keep running, never sleeping, if you will, because HDD lifespans are simply not long enough. So it will keep running and continue to waste energy. It’s worth pointing out that building HDDs is also a high-energy process.
“Considering the life span of HDD is about four or five years, and SSD is ten years more, so the whole industry needs to work out a new technology or storage media to accommodate the increase in warm data,” Zhou says. “I keep my eye on these developments because, in my opinion, HDD is not the final answer.”
For now, Huawei is answering the demands of current data storage needs with AI-ready data storage. These include the New-Gen OceanStor Dorado Converged All-Flash Storage for mission-critical production workloads and OceanStor A Series High-Performance AI Storage for AI Model Training and Inference.
The Huawei OceanStor Pacific All-Flash Scale-Out Storage, which is for mass data storage, provides industry-high density and low power consumption. With an exabyte level (equal to 1,000 petabytes) of scalability. For data protection, Huawei has also launched its OceanProtect All-Flash Backup Storage, which promises five times faster data recovery than industry alternatives.
For more details about Huawei’s data storage offerings, visit: https://e.huawei.com/en/products/storage
