FEATUREDTechnology

Huawei’s new NAS solution wants to tackle the biggest challenges in GenAI


When it comes to data storage, it’s fair to say AI has had a destabilizing effect on the industry as a whole, as businesses looking to develop and deploy the technology are faced with unfathomably large amounts of data and a dizzy number of files. 

It’s a problem Huawei is looking to solve with its high-performance OceanStor A800 network storage attached (NAS) device, first revealed in 2023, and this week getting a European outing at the company’s Innovative Data Infrastructure (IDI) Forum 2024 in Berlin. 

Like Huawei’s Dorado and Pacific NAS devices, the focus here is on AI data storage. Huawei calls it the “new data paradigm,” and an “accelerated data awakening.” Building, processing generative AI training models from scratch, maintaining that all-important data sovereignty, ensuring system reliability, it all demands a bafflingly large amount of space and speed. That’s a big problem for big businesses – and that’s where this uncompromising black box fits into the array.   

Huawei's OceanStor A800 NAS on show at IDI 2024

(Image credit: Huawei / Future)

Speed, space, and ‘awakening data’

The OceanStor A800 fits into an admittedly niche enterprise market. We’re not talking storage for the masses here, however nice 500GB a second bandwidth would be. According to Huawei, the A800 is capable of 24 million IOPS per controller enclosure, delivering ten-times the performance of existing storage, ten-times the data mobility. The storage also supports bandwidth in PB/s and 100 million IOPS, and boasts data reliability of 99.9999%. 

Huawei's OceanStor A800 stats in a slide at IDI 2024

(Image credit: Huawei / Future)

At IDI 2024, the company stated the A800 “can increase AI cluster utilization by 30%, and delivers high bandwidth and IOPS, which are four and eight times better than its peer vendor. [The device] supports scaling out to EB-level capacity with up to 512 controllers, as well as scaling up to a maximum of 4,096 computing cards.”



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *