Managed Lustre

Google Cloud introduced a multi-node KV cache offloading solution using GKE and Managed Lustre for large language model inference, achieving over 50% TCO savings and nearly 60% reduction in GPU-hour r…