Pallas: A Proactive KV Cache Migration Framework for LLM Inference in AI-RAN
2026-08-17 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address a problem in mobile AI services where a phone moving between cell towers can lose connection to its ongoing AI task. They propose Pallas, a system that prepares the necessary AI data ahead of time at the next tower based on predicting user movement, so the AI can smoothly continue without long delays. This method splits the data into a stable part and a changing part, allowing the next tower to get most information early and finish the rest quickly after handover. Their tests show this approach significantly reduces service interruptions and speeds up response times compared to existing methods.
Large Language Model (LLM)Cellular HandoverKey-Value (KV) CacheInference StateInter-Token Latency (ITL)Service Interruption Time (SIT)PrefetchingMobility PredictionBase Station (gNB)
Authors
Tianhang Ding, Jianchun Liu, Hongli Xu
Abstract
AI-RAN brings large language model (LLM) serving close to mobile users, but cellular handover can separate an active request from its inference state: the user attaches to a target base station (gNB) while the large and growing key-value (KV) cache remains at the source. Retaining inference at the source preserves service continuity but persistently increases inter-token latency (ITL), whereas recovering the state at the target restores serving locality but requires KV-cache transfer, recomputation, or a combination of both only after handover, directly prolonging service interruption time (SIT). This work presents Pallas, a \textit{proactive} KV-cache migration framework that prepares the inference state at the predicted target before handover, in parallel with ongoing source-side inference and token delivery. At the preparation trigger, Pallas partitions the token sequence into a stable historical prefix and an evolving suffix. The target reconstructs the prefix through local prefill, while the source streams the KV blocks generated for the suffix. At handover, the target assembles both portions into an up-to-date KV cache and resumes decoding locally, leaving only unfinished preparation to contribute to SIT. An online scheduler selects the \textit{prefetching window}, which determines how early preparation begins before handover, based on mobility predictions and runtime telemetry. Across three LLMs and $100$--$500~\mathrm{Mbps}$ inter-gNB links, our vLLM-based prototype reduces average SIT by factors of $2.28$--$89.68$ over target-side recovery approaches and lowers average ITL by $16.0\%$--$50.0\%$ compared with source-side forwarding.