Everpure FlashBlade//S500 solutions for NCP deployments are based on a master building block of a single 10 chassis Flashblade//S500 per 1,024 GPUs. This base building block ensures that performance consistency is maintained as the NCP cluster size grows. The recommended scaling model involves adding a FlashBlade//S500 incrementally per 1,024 GPUs.
Each FlashBlade//S500 for NCP is fully capable of performing in both single-tenant and multitenant configurations at 1024 GPUs, maintaining high performance for 1024 GPU read and write benchmarks in both scenarios
With clientless native NFSoRDMA support, FlashBlade//S500 provides simultaneous data access to thousands of data connections with minimal administrative overhead.
All data on a FlashBlade//S500 is accessible via single data network IP.