At the Huawei Innovative Data Infrastructure Forum 2024 held in Berlin, Dr. Peter Zhou, Huawei Vice President and President of Huawei Data Storage Product Line, gave a talk titled “Redefining Data Storage in the Data Awakening Era”. He introduced Huawei's latest thinking.
AI is disrupting traditional data storage by focusing not only on performance, reliability, and data paradigms, but also on scalability, sustainability, and data fabric, Zhou said. He added that in the era of data awakening, Huawei will redefine data storage through cutting-edge innovations in six aspects:
- Ultra performance: Huawei enhances storage performance by 10 times compared to traditional storage. This storage also supports PB/s bandwidth and 100 million IOPS, greatly increasing the efficiency of the entire generative AI process.
- Data resiliency: Innovative architecture and technology ensures high reliability of 99.9999%. Built-in ransomware detection engine increases detection accuracy to his 99.99%. Checkpoint recovery time during AI training has also been reduced to less than 1 minute.
- New data paradigm: Multidimensional tensor data is enabled to support fast data retrieval by intelligent search engines. Search Augmentation Generation (RAG) technology works with embedded knowledge bases to eliminate hallucinations in large-scale AI models.
- Scalability: You can scale out a single storage cluster to EB-level capacity, and each engine can be scaled up with more GPUs, DPUs, or NPUs for near-storage compute.
- sustainability: Innovations in storage media and devices have enabled superior storage energy efficiency (less than 1 Watt/TB) and storage density (greater than 1 PB/U).
- Data fabric: Storage metadata management and search capabilities provide global data visibility and manageability, and 10x more efficient data mobility.
These six aspects guide the philosophy behind the recent launch of the high-performance OceanStor A800, a powerful addition to Huawei OceanStor A-series storage models. Tuned for AI applications, OceanStor A800 can improve AI cluster utilization by 30%, and when it comes to performance, it delivers high bandwidth and IOPS that are 4x and 8x better than competing vendors. Masu. In terms of scalability, OceanStor A800 supports scale-out to EB-level capacity with up to 512 controllers and scale-up to up to 4,096 compute cards. When it comes to space and energy savings, it delivers an impressive storage density of 1 PB/U and energy efficiency of 0.7 Watt/TB. It also offers new data paradigms with vector indexing, tensor data, and RAGs. In terms of data resiliency, ransomware detection accuracy has increased from 99.9% to 99.99%. In addition, data fabric capabilities make it easier to manage your data assets.
At the same time, innovations in storage media are driving sustainable development. Huawei's newly released high-capacity SSD can provide 10 times more capacity with the same disk size, further reducing data center energy consumption. With a capacity of 128 TB per disk, the new SSD takes up 88% less storage space and consumes 92% less energy than its peer vendor's SSD when storing 1 PB of data.
To be AI-ready, companies need to prepare their data. The Omni-Dataverse global file system built into DME makes enterprise data assets visible, manageable, and mobile across geographies, creating a solid AI data lake storage foundation for enterprises.
Dr. Peter Zhou concluded with Huawei's commitment to redefine data storage and build a cutting-edge AI-enabled data infrastructure to enhance customer value, focusing on customer challenges and demands in the data-aware era. He concluded by emphasizing.