Arrow Electronics, a global technology provider, announced a new collaboration with PEAK:AIO, in which Arrow's intelligent solutions division will provide the integrated hardware platform, enabling resellers and solution providers to offer high-performance AI storage devices as part of GPU infrastructure deployments.
The solution is designed to enable organizations managing modern AI workloads to achieve high throughput and reliable, low-latency storage performance, without the cost and operational complexity of traditional enterprise storage or high-performance computing (HPC) parallel file systems.
PEAK:AIO's AI Data Server is designed to adapt to the effective scalability of artificial intelligence applications. It starts with a single 2U server, expandable on a modular basis as your needs grow, and can reach exabyte levels efficiently, without sacrificing the simplicity of a standard Linux NFS environment. The solution was specifically developed for small to medium-sized GPU clusters, not for legacy CPU-based HPC superclusters. The PEAK:AIO platform is designed to constantly feed GPUs with data, maximizing utilization without excessive investment in storage infrastructure.
Tonya Roach, vice president of sales for Arrow Intelligent Solutions Business, said: “AI infrastructure is evolving rapidly, and channel partners are being asked to deliver performance at scale without increasing complexity. PEAK:AIO offers a clear and pragmatic approach to AI storage, and Arrow is uniquely positioned with our strong integration capabilities and supply chain.”
Roger Cummings, CEO and president of PEAK:AIO, added: "AI teams demand storage that keeps their GPUs busy, scales easily, and doesn't lock them into proprietary stacks. In today's market, the fact that PEAK:AIO can do all this with less hardware than competing solutions isn't just about performance, it's about sourcing. Our partnership with Arrow gives us the global reach to provide these services wherever customers are building their solutions."
I PEAK:AIO AI Data Server e Open pNFS sono progettati per essere compatibili con qualsiasi fornitore di server e con una ampia gamma di piattaforme hardware, aiutando i clienti a non rimanere vincolati a stack client proprietari e a evitare il vendor lock-in.
Le aziende possono rivolgersi ai propri distributori autorizzati per acquistare la soluzione.
