Platform for AI (PAI) - KV Store Global Context Cache Release
Dec 19 2025
Platform for AI (PAI)When you deploy the LLM service in EAS, you can configure the global online context cache of KV Store to improve the LLM inference throughput.
Content
Optimization: EAS supports configuring the global online context cache of KV Store in the process of deploying LLM service, and improves kvcache hit rate and LLM inference throughput performance through multi-level storage query of GPU → memory → Redis kv metadata.
Help Document
https://alibabacloud.com/help/doc-detail/110985.htm