The Gulf Time Newspaper One of the finest business newspapers in the UAE brought to you by our professional writers and editors.
ABU DHABI/GULF TIME
VDURA is expanding its presence in the UAE and wider Middle East as investment in artificial intelligence and digital transformation accelerates across the region. Speaking at Ai Everything Abu Dhabi, Matt Swalley, Senior Director at VDURA, said the company was excited to establish a stronger presence in the region, describing the exhibition as its first major event in the market.
VDURA provides high-performance data storage infrastructure for AI workloads. Swalley said the company has a team in Dubai and is holding discussions with local accelerators, government organisations and other stakeholders.
VDURA also featured its partner Hafþór “Thor” Björnsson at the event. Björnsson is known for playing The Mountain in Game of Thrones. According to Swalley, the priorities of AI infrastructure providers have changed significantly as AI deployments have moved from experimentation toward larger-scale production.
The focus is no longer simply on deploying more GPUs, he said. The challenge is increasingly about ensuring that those GPUs can be used efficiently and that organisations can generate a return from the infrastructure they have already deployed.
Storage plays an important role in that process because AI systems depend on a continuous flow of data between compute resources and storage. For training workloads, data needs to be accessed quickly, while checkpointing requires large amounts of information to be written to storage. During checkpointing, GPUs can become idle while the data is being saved, reducing the utilisation of expensive computing resources.
Inference presents another challenge. As AI applications work with increasingly large amounts of context, storage needs to provide fast access to information beyond the data immediately available in GPU memory. Swalley said these requirements are driving greater attention to how quickly data can move between compute and storage.
VDURA’s Data Platform V12 is designed to address these storage requirements through its parallel file system architecture. Rather than relying on a single path between compute and storage, the architecture allows data to move in parallel across multiple storage nodes. The approach is designed to support the high-performance requirements of AI workloads, particularly as GPU infrastructure continues to scale.
V12 also combines flash and hard-disk storage within a single platform. Data initially lands on flash, while VDURA’s Context-Aware Tiering is designed to move data between flash and hard disks as access patterns change. This allows operators to use flash for performance-sensitive workloads while using higher-capacity disk storage for data that is accessed less frequently.
The approach is intended to give AI infrastructure providers a way to balance performance and capacity without maintaining separate storage platforms. Swalley said VDURA sees significant potential in the UAE and the wider Middle East. He pointed to the region’s investment in digital transformation and AI, as well as its resources and expertise, as factors contributing to its rapid development.
For VDURA, that creates an opportunity to expand its presence and work with organisations building AI infrastructure in the region. The company is relatively new to the market, Swalley said, but has an experienced team in place and is engaged in discussions across government and AI cloud environments.
Another part of VDURA’s approach is its open-hardware model. Swalley said the company works with hardware and technology partners so that its software can operate across different infrastructure components. He cited partnerships involving companies such as Dell, Supermicro and AMD. The approach is intended to give customers greater flexibility when designing AI infrastructure and to make it easier to work with their preferred technology and supply-chain partners.
Swalley said VDURA is also having discussions with partners about jointly taking solutions to enterprise customers in the region. Looking ahead, Swalley said one of the biggest challenges for AI infrastructure providers will be dealing with the speed at which the technology and its requirements are changing. The future requirements of AI workloads are difficult to predict, he said, making flexibility an important consideration when organisations design their storage infrastructure.
VDURA’s approach is to allow customers to expand different parts of the storage environment as their requirements change. The platform can independently scale performance, capacity and metadata resources, giving operators the ability to adapt their infrastructure rather than being tied to a fixed configuration.
For AI infrastructure providers facing rapidly changing workloads, Swalley said the goal is to provide storage that can evolve alongside the technology.