Huawei and Nvidia Move AI KV Cache to Dedicated Storage; SSD Standard Still Undefined
Overview
Huawei's OceanStor M900 and Nvidia's CMX Context Memory Storage both place reusable AI inference KV Cache in a shared, tiered storage layer, so cache hits reduce repeated computation.
Industry sources interviewed by Leiphone (雷峰网) say no agreed SSD specification yet exists for this workload, with endurance requirements the main point of disagreement. One storage executive said the endurance target for one design rose from 3 to 9 drive writes per day and could change again.
Sources expect the industry to take 6 to 12 months to converge on a common standard, followed by roughly another year of development, validation and tuning. Pandaily's later report repeats the same figures, but neither report itemizes its underlying evidence, so the specific endurance numbers rest on the sourcing alone.
Written by AI from the articles below · updated Oct 8, 10:58 PM ET
Check the sources:
Article timeline
Follow the coverage from different perspectives. Times are ET.
- PandailyHuawei's KV cache storage faces a missing SSD endurance standard
AIHuawei's OceanStor M900 and Nvidia's CMX move reusable inference KV cache into a shared storage tier, but no agreed SSD specification exists for it. A storage executive said endurance requirements for one design rose from 3 to 9 drive writes per day and could change again. Industry sources expect convergence to take 6 to 12 months, with another year for development and validation.
- LeiphoneHuawei and Nvidia Move KV Cache Into Dedicated Storage, but SSD Specs Remain Undefined
AIHuawei launched the OceanStor M900 and Nvidia introduced CMX Context Memory Storage to manage inference KV Cache in tiered storage, reducing repeated computation on cache hits. Sources say SSD requirements for this workload are not standardized: some plans have moved from 3 DWPD to 9 DWPD, and the industry may need 6–12 months to reach a common product standard.
Heat trend
Not enough continuous observations to show a trend yet.