From 3d109d50be3f345ccade155b848a81c8aa65eeff Mon Sep 17 00:00:00 2001 From: "minke.yu" Date: Fri, 9 Oct 2026 19:30:53 +0800 Subject: [PATCH] =?UTF-8?q?fix:=20l3test=20L2=20=E6=94=B9=20ratio=201.0?= =?UTF-8?q?=EF=BC=880.05=20=E4=BC=9A=E5=99=8E=E6=AD=BB=20L3=20=E5=86=99?= =?UTF-8?q?=E5=85=A5=E8=B7=AF=E5=BE=84=EF=BC=89?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- deploy/b300-ds41/dockerserve-3p1d-l3test.yml | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/deploy/b300-ds41/dockerserve-3p1d-l3test.yml b/deploy/b300-ds41/dockerserve-3p1d-l3test.yml index 2014b48cc1..ad8cfd07ad 100644 --- a/deploy/b300-ds41/dockerserve-3p1d-l3test.yml +++ b/deploy/b300-ds41/dockerserve-3p1d-l3test.yml @@ -1,6 +1,6 @@ # 【L3 cache-hit 测试变体 2026-10-09】基于 dockerserve-3p1d-otlp.yml,仅改 P: -# --max-total-tokens 800000(L1 缩到 ~0.8M token)+ --hicache-ratio 0.05(L2 缩到 ~4 万 token;V4 hicache 不支持 --hicache-size) -# 使 ~1.7M token/ P 的语料溢出 L1+L2,热轮强制走 L3 mooncake 命中。 +# --max-total-tokens 800000(L1 缩到 ~0.8M token)+ --hicache-ratio 1.0(L2=L1=0.8M token。⚠️ 教训:ratio 0.05 时 L2 太小,build_storage_backup_spec 拿不到 host 页会静默跳过 L3 写入,热轮无数据可中;V4 hicache 不支持 --hicache-size) +# 使 ~2.6M token/P 的语料(48×160k)溢出 L1+L2(1.6M),热轮强制走 L3 mooncake 命中。 # 与主模板同容器名/端口,用前必须先 down 主栈。D 与 router 不变。 # 原注释: 3x cp2-P + 1x dp2-D(cg256) PD 分离 + L3 mooncake + engram + OTLP trace(2026-09-27) # 基于 dockerserve-3p1d-soak.yml 增加: --enable-trace --trace-modules request,mooncake @@ -67,7 +67,7 @@ services: --enable-cache-report --enable-metrics --speculative-algorithm DSPARK --speculative-dspark-block-size 5 --enable-hierarchical-cache - --hicache-ratio 0.05 + --hicache-ratio 1.0 --hicache-mem-layout page_first_direct --hicache-io-backend direct --hicache-write-policy write_through @@ -107,7 +107,7 @@ services: --enable-cache-report --enable-metrics --speculative-algorithm DSPARK --speculative-dspark-block-size 5 --enable-hierarchical-cache - --hicache-ratio 0.05 + --hicache-ratio 1.0 --hicache-mem-layout page_first_direct --hicache-io-backend direct --hicache-write-policy write_through @@ -147,7 +147,7 @@ services: --enable-cache-report --enable-metrics --speculative-algorithm DSPARK --speculative-dspark-block-size 5 --enable-hierarchical-cache - --hicache-ratio 0.05 + --hicache-ratio 1.0 --hicache-mem-layout page_first_direct --hicache-io-backend direct --hicache-write-policy write_through