1. 09 Apr, 2026 1 commit
  2. 08 Apr, 2026 1 commit
  3. 18 Mar, 2026 1 commit
  4. 13 Mar, 2026 1 commit
  5. 10 Mar, 2026 1 commit
  6. 09 Mar, 2026 1 commit
  7. 06 Mar, 2026 1 commit
  8. 03 Mar, 2026 1 commit
  9. 28 Feb, 2026 1 commit
  10. 13 Feb, 2026 1 commit
  11. 07 Feb, 2026 1 commit
  12. 05 Feb, 2026 1 commit
  13. 04 Feb, 2026 1 commit
    • zhanqiuhu's avatar
      [Metrics] Add labeled prompt token metrics for P/D disaggregation (#33290) · 4403e3ed
      zhanqiuhu authored
      
      
      Add labeled Prometheus metrics to distinguish where prompt tokens come
      from in P/D disaggregated deployments.
      
      In P/D disaggregation, decode instances receive KV cache from prefill instances.
      Currently, decode reports inflated prompt throughput because it counts all
      prompt tokens as "computed", even though most were transferred.
      
      This PR adds labeled metrics so users can understand actual compute work vs
      transferred work:
      
      vllm:prompt_tokens_by_source_total{source="local_compute"}        # Tokens prefilled locally
      vllm:prompt_tokens_by_source_total{source="external_kv_transfer"} # Tokens received via KV transfer  
      vllm:prompt_tokens_by_source_total{source="local_cache_hit"}      # Tokens from local prefix cache
      vllm:prompt_tokens_cached_total                                    # Total cached (local + external, -1 when all 
      Signed-off-by: default avatarZhanqiu Hu <zh338@cornell.edu>
      4403e3ed
  14. 24 Jan, 2026 1 commit
  15. 12 Jan, 2026 1 commit
  16. 23 Dec, 2025 1 commit
  17. 10 Dec, 2025 1 commit
  18. 25 Nov, 2025 1 commit
  19. 14 Nov, 2025 1 commit
  20. 05 Nov, 2025 1 commit
  21. 12 Oct, 2025 1 commit
  22. 05 Oct, 2025 1 commit
  23. 19 Sep, 2025 1 commit
  24. 12 Sep, 2025 1 commit
  25. 29 Aug, 2025 1 commit
  26. 16 Aug, 2025 1 commit
  27. 13 Aug, 2025 1 commit
  28. 30 Jul, 2025 1 commit
  29. 19 Jul, 2025 1 commit
  30. 23 Jun, 2025 1 commit
  31. 19 Jun, 2025 1 commit
  32. 04 Jun, 2025 1 commit
  33. 03 Jun, 2025 1 commit
  34. 30 May, 2025 1 commit
  35. 23 May, 2025 1 commit
  36. 12 May, 2025 1 commit
  37. 30 Apr, 2025 1 commit
  38. 26 Apr, 2025 1 commit
  39. 23 Apr, 2025 1 commit
  40. 17 Apr, 2025 1 commit