🆕 🚀 We're excited to announce Qumulo Cloud AI Accelerator: a faster way for enterprises to run AI workloads wherever GPU capacity is available, without copying massive datasets to get there. GPU scarcity is only half the problem. The other half is data gravity. When GPUs finally become available in a new region or zone, the data is never where they are. Teams end up paying for idle compute while hundreds of terabytes move into position — or maintaining expensive replicated copies across every location just in case. Either way, the meter runs before any real work begins. Qumulo Cloud AI Accelerator eliminates that tax. Combined with Cloud Data Fabric and NeuralCache, it presents any enterprise dataset in real time to any GPU farm, in any cloud, without replication or pre-staging. GPU hunting becomes a scheduling decision, not a logistics problem. And with Cisco UCS and networking providing the enterprise infrastructure foundation, the full stack is built for the scale and reliability AI workloads demand. Read the full announcement ➡️ https://bwnews.pr/4uyj8r7 We'll be at #CiscoLive in Las Vegas starting May 31. Come find us and see what we're building together.
Joseph Huff
3d
Your data is one of your most valuable assets — so why create unnecessary copies, replicas, and storage silos just to make it accessible? With Qumulo, organizations can project data where it’s needed, when it’s needed, without the operational complexity and cost of managing duplicate datasets. Reduce storage sprawl. Simplify data access. Enable your teams and applications to work from a unified data platform across edge, core, and cloud environments. The future isn’t about managing storage. It’s about maximizing the value of your data.
Hooman Parta
5d
Douglas Gourlay This is the storage world finally catching up to what inference economics has been screaming for two years: compute and data both have to become fungible, not just one of them. Everyone optimized the last few inches between flash and GPU — almost no one attacked the data-gravity tax that decides whether those GPUs ever get fed in the first place. Decoupling the two, and letting workloads chase capacity across clouds and regions in minutes, is the real unlock. Strong launch.
So uh... just connecting all your data in a single namespace and letting you use it the way your business wants to use it... jus' sayin'...
James Pepin
5d
Sounds like 1980s 90s to feed data to national hpc centers and or other stupid net tricks that followed on from then. What’s new is old again
Ala Sajed
5d
A solution for all customers to take advantage of
Every AI team I've talked to in the last year has hit the same wall. They finally get GPU capacity allocated, then spend days or weeks moving data before a single training run starts. The meter is running the whole time. That's what Cloud AI Accelerator is built to fix. Real-time access to enterprise datasets from any GPU farm, in any cloud, without replication or pre-staging. Combined with Cisco UCS underneath, it's an honest answer to what customers have been asking for. Any Data in Any Location or Cloud, without ever losing control or consistency. Looking forward to many conversations next week.