# Scheduling **URL:** https://discuss.vllm.ai/c/features/scheduling/24.md [Latest](https://discuss.vllm.ai/latest.md) · [Categories](https://discuss.vllm.ai/categories.md) --- ## [About the Scheduling category](https://discuss.vllm.ai/t/about-the-scheduling-category/141)
**Author:** [@comaniac](https://discuss.vllm.ai/u/comaniac)\ **Replies:** 0\ **Last updated:** [March 20, 2025, 5:14pm UTC](https://discuss.vllm.ai/t/about-the-scheduling-category/141 "2025-03-20T17:14:23Z")
Any topics relate to scheduling. For examples: Scheduling policies, Scheduler behaviors, Pluggable scheduler, etc --- ## [Why vllm does not support LMP?](https://discuss.vllm.ai/t/why-vllm-does-not-support-lmp/1763)
**Author:** [@Binwon](https://discuss.vllm.ai/u/Binwon)\ **Replies:** 3\ **Last updated:** [October 23, 2025, 9:33am UTC](https://discuss.vllm.ai/t/why-vllm-does-not-support-lmp/1763 "2025-10-23T09:33:17Z")
I found that Longest Prefix Matching (LPM) is used in SGLang as part of its cache-aware scheduling. I also tested a similar implementation in vLLM, and it slightly improved throughput. Has anyone else tried implementin… --- ## [Is FCFS Scheduling Holding Back vLLm's Performance in Production?](https://discuss.vllm.ai/t/is-fcfs-scheduling-holding-back-vllms-performance-in-production/1584)
**Author:** [@sidikbro](https://discuss.vllm.ai/u/sidikbro)\ **Replies:** 3\ **Last updated:** [September 11, 2025, 9:46am UTC](https://discuss.vllm.ai/t/is-fcfs-scheduling-holding-back-vllms-performance-in-production/1584 "2025-09-11T09:46:35Z")
Hi vLLm community :waving\_hand: We’ve been exploring scheduling strategies to better support mixed workloads. We’ve observed a challenge similar to a supermarket checkout: when short/interactive requests (buying one ite… --- ## [Why is cuda graph capture sizes limited by max\_num\_seqs](https://discuss.vllm.ai/t/why-is-cuda-graph-capture-sizes-limited-by-max-num-seqs/954)
**Author:** [@gautijha37](https://discuss.vllm.ai/u/gautijha37)\ **Replies:** 1\ **Last updated:** [June 29, 2025, 5:32pm UTC](https://discuss.vllm.ai/t/why-is-cuda-graph-capture-sizes-limited-by-max-num-seqs/954 "2025-06-29T17:32:16Z")
My understanding for forward pass in vllm is that it is done on a batch of tokens taken from the scheduler output and limited by max\_num\_batched\_tokens. In the forward pass I see these number of tokens are padded to the … --- ## [Why does computation time remain consistent across chunks in chunked-prefill despite linearly growing attention complexity?](https://discuss.vllm.ai/t/why-does-computation-time-remain-consistent-across-chunks-in-chunked-prefill-despite-linearly-growing-attention-complexity/743)
**Author:** [@chj](https://discuss.vllm.ai/u/chj)\ **Replies:** 3\ **Last updated:** [June 2, 2025, 9:43am UTC](https://discuss.vllm.ai/t/why-does-computation-time-remain-consistent-across-chunks-in-chunked-prefill-despite-linearly-growing-attention-complexity/743 "2025-06-02T09:43:38Z")
Question: Why does computation time remain consistent across chunks in chunked-prefill despite linearly growing attention complexity? Context: I’m profiling vLLM’s chunked-prefill optimization for long sequences (e.g., … --- ## [Does vLLM support multiple model\_executor?](https://discuss.vllm.ai/t/does-vllm-support-multiple-model-executor/541)
**Author:** [@leeeizhang](https://discuss.vllm.ai/u/leeeizhang)\ **Replies:** 1\ **Last updated:** [April 28, 2025, 9:53am UTC](https://discuss.vllm.ai/t/does-vllm-support-multiple-model-executor/541 "2025-04-28T09:53:38Z")
Hi, all. My model is relatively small-scale and doesn’t require large tensor parallelism (TP) for inference (TP=2 is enough for this model). However, my hardware instance has 8 GPUs. How can I configure multiple executo… --- ## [V1 Chunked Prefill Scheduling Policy: how prefill would be scheduled?](https://discuss.vllm.ai/t/v1-chunked-prefill-scheduling-policy-how-prefill-would-be-scheduled/146)
**Author:** [@liuzijing2014](https://discuss.vllm.ai/u/liuzijing2014)\ **Replies:** 8\ **Last updated:** [March 25, 2025, 8:03am UTC](https://discuss.vllm.ai/t/v1-chunked-prefill-scheduling-policy-how-prefill-would-be-scheduled/146 "2025-03-25T08:03:35Z")
Hi guys, some question about V1 chunked prefils scheduling policy: it seems that it would prioritize decode at first. After we schedule as much as “decode”, how do we determine which prefill to be scheduled first? Is i…