H3C releases next-generation AI data storage platform

In the process of AI evolution, data elements play a crucial role. Data not only provides massive learning resources for artificial intelligence, but also promotes the cross-border integration of artificial intelligence applications, leading the innovation and upgrading of artificial intelligence algorithms. With the rapid development of AI, the amount of data is showing explosive growth, and storage systems are facing three major challenges: increasing model parameters, complex data types, and uninterrupted business operations. To achieve large-scale intelligent productivity, it is necessary to walk on two legs: storage power and computing power. As a leader in digital solutions, New H3C Group closely follows market demand, and through comprehensive restructuring of systems, experiences, and values, creates a new generation of storage platforms to help industry customers cope with storage challenges with ease.

The H3C UniStor Polaris X20000 series is a fully restructured distributed storage software platform that enhances system performance through intelligent optimization and streamlining of IO data streams, achieving performance optimization, protocol fusion, and architecture evolution.

• Performance optimization: Intelligent optimization of IO read and write paths; Zero copy memory reduction with invalid IO; High performance EPC client, IO level load balancing for all nodes, supporting parallel file systems; CPU intelligent grouping, maximizing CPU utilization.
• Protocol Fusion: A system that simultaneously satisfies block, file, object, and HDFS applications; Cross protocol access performance is lossless; Unified logical resource pool to achieve semantic lossless protocol fusion.
• Availability improvement: fully distributed multi active architecture, with fault switching in seconds; Fault MS level detection and isolation; Support multi tenant data and resource isolation.

Based on the underlying innovation of system refactoring, the H3C UniStor Polaris X20000 series will bring customers a stronger performance experience, smarter management experience, and more efficient availability experience, providing a smart storage engine for enterprise digital transformation.

• Refactoring performance experience maximizes the release of GPU effective computing power: node bandwidth reaches 80GB/s, 2 million IOPS, and large-scale cluster performance shows near linear growth, which can maximize the release of GPU effective computing power.
• Refactoring the management experience with a resource pool that supports full process intelligent computing applications: a single system can simultaneously support lossless access to files, objects, and big data across protocols without the need for data migration, and the available capacity space is greatly increased.
• Refactoring the availability experience to ensure uninterrupted business switching in seconds: significantly improving business recovery speed; Multi tenant independent capacity quota, performance QoS guarantee, without affecting each other; Support online hardware replacement, zero downtime to ensure uninterrupted business operations; Support one-time expansion of multiple hosts, significantly reducing business impact.

• AIGC Text Large Model Application Fields: Supports 70B parameter level large model training, improves Checkpoint loading speed by 50%, and significantly shortens training cycles.
• In the field of gene sequencing: the comparison cycle of gene sequencing analysis has been shortened by 57%, from 30 hours to 13 hours.
• Automatic driving field: the development iteration cycle of the auto drive system was shortened by 55%, from an average of 26 months to 13 months, significantly improving the production efficiency of the enterprise.

As the data foundation for the overall solution of New H3C Group’s intelligent computing, the H3C UniStor Polaris X20000 series not only supports unified scheduling and system management of the Aofei computing platform, but also provides business level reliability, intelligent IO path optimization, data global control, and data security strategies for upper level AI applications, widely empowering new data storage scenarios such as AIGC, smart factories, financial imaging, satellite surveying, etc., and creating a solid data foundation for the development of intelligent computing.