iCreatAI News
Skip to content
Together AI Blog·· 2 days agoAI score55

Introducing Together Storage Accelerator: fast, free data caching for your GPU cluster

Introducing Together Storage Accelerator: fast, free data caching for your GPU cluster

AI summary

Together AI launched Together Storage Accelerator (TSA), a distributed caching layer included free with every Together GPU Cluster that caches object storage and Hugging Face Hub reads on cluster NVMe, presenting drop-in S3-compatible and HF endpoints without application changes. Together reports ~20× faster warm reads (up to 3.4 GB/s per node) versus cross-cloud origins, ~60–80× lower latency on small files, and one origin fetch per cluster for a 70 GB checkpoint pulled by 200 nodes. TSA is powered by ASTRA and pairs with Together's T4 object storage, currently in limited preview.

Source: Together AI Blog · together.ai