Add files using upload-large-folder tool

AUDIT_20AGENT_REPORT.md

@@ -0,0 +1,187 @@
+# 20-Agent Deep Audit Report
+
+Date: 2026-04-19 21:55  
+Scope: Method "Stall-Aware Embedding Layout + E+L0 Dual-Path Routing + Host LFU L0"  
+Wave 1: 12 agents（narrative / prior art / code / fairness）  
+Wave 2: 8 agents（statistics / simulation / data / traceability）
+
+---
+
+## 💀 論文不可送審的 Catastrophic Blockers（需立刻修）
+
+### 1. RESULTS.md Optane row 數字錯 → 結論 INVERT
+- Claim "MaxEmbed 5.62 ms/batch, CV 5.4%, 贏 E+L0 2.31x"
+- **真實**：MaxEmbed overall_us = **15.80 ms/batch, CV 21.3%**
+- **E+L0 在 Optane 也贏 9.5x，不是輸 2.31x**
+- Agent 13: parse unit swap（5.62 是 per-sample μs，5.4% 是 ops/s CV）
+- **立刻修 RESULTS.md + 所有引用該結論的地方**
+
+### 2. Paper 捏造 stall + scheduling 實驗數字
+- Agent 3 + 20 雙重確認：
+  - 所有 stall numbers（80.10 / 11.47 / 6.02 / 4.04 / 5.08 / 7.80 / 10.38 / 14.41 / 3844 ms）都是 **analytic 推導**，但 paper 寫成 "MQSim cycle-accurate 輸出"
+  - 隱含 per-miss latency 在不同 step 不一致（7.56/37/47/9.34μs 不可能同時為真）
+  - **Tagged flash 234,895 完全虛構**（實際 92,331）
+  - **Scheduling flash 5,160/5,224/26,132/31,600 虛構**（沒跑過 miss_first/hit_first/spread_miss 實驗）
+  - SE spread-miss 29,532 **MISLABELED**（實際是 cog_v2_se K10000）
+- **行動**：重跑 scheduling 實驗或**改寫 paper 去掉假數字**
+
+### 3. Paper Narrative 3 docs 互相矛盾 + E+L0 核心未寫
+- Agent 1 + 19 雙重確認：
+  - `paper_story.md` 推 miss-first
+  - `algorithm_design.md` 推 COG v2 spread-miss
+  - `page_oriented_embedding_spec.md` 是舊 project
+  - **E+L0 Dual-Path Routing + Host LFU L0（宣稱的核心）3 個 doc 都沒寫**
+- `dual_path_pollution_model.md` 是 **100% vapor claim**：Score_MEM / PC(p) / ConflictRisk / 最佳 split k* / frequency-sketch admission filter **全無 code 實作**，code 用 hardcoded EWMA_THETA=0.926f（archived doc previously labelled the admission filter "TinyLFU" — that claim is now withdrawn; the shipped L0 is aging LFU, Arlitt et al. 2000）
+- **行動**：完全重寫 paper_story + algorithm_design，用 E+L0 + Host LFU L0 當核心
+
+### 4. MaxEmbed Tail 比較不成立
+- Agent 14：MaxEmbed 報的是 **windowed avg 100-200 batches**（CLT 壓低 variance ~10x）
+- 我方 E+L0 是 true per-batch
+- Paper claim "MaxEmbed tail 窄 p99/mean=1.04" 是 **averaging artifact**
+- 真實 MaxEmbed tail 未知
+- `batch_latencies.csv` 根本沒存，每 run 被覆蓋
+- **行動**：**不能 claim tail**，或重跑兩邊都 dump per-batch CSV
+
+### 5. Dataset Scale 兩個 dataset 不對
+- CriteoTB: 宣稱 882M/4.37B，實際 126.6M/450M（-85% / -90%），HF 10% subsample
+- Alibaba-iFashion: 宣稱 4.46M items，實際 2.18M（-51%）referenced in 999K outfits
+- **行動**：修 paper wording 或下 Tianchi 370GB full
+
+---
+
+## 🔴 Code Bugs（修不難，立刻動）
+
+| # | Bug | File:Line | Agent | Severity |
+|---|-----|-----------|-------|:--------:|
+| 6 | Mode 2 timer 包 `posix_memalign` + `pthread_create` | `race_trace_v3.c:580` | 7 | 🔴 |
+| 7 | `nhit = nm` Mode 2 永遠 report miss=0，CSV 說謊 | `race_trace_v3.c:589` | 7 | 🔴 |
+| 8 | Mode 2 無 `_mm_lfence`（Mode 4 有）→ OoO overlap 讓 Mode 2 假快 | `race_trace_v3.c` | 7 | 🔴 |
+| 9 | `l0_hit_primary_or_replica` 定義但 unused（replicas dead code） | `race_trace_v3.c` | 7 | 🔴 |
+| 10 | LFU tie-break 永遠 bias way-0（comment 說 xorshift 但沒實作） | `race_trace_v3.c:252-258` | 8 | 🔴 |
+| 11 | LFU hash `pg % sets` 非均勻（table_offset bit clustering） | `race_trace_v3.c:231` | 8 | 🔴 |
+| 12 | LFU 16-bit counter 無 aging → scan resistance failure | `race_trace_v3.c:254` | 8 | 🟡 |
+| 13 | Criteo encoding insertion-order（paper 是 descending-freq） | `criteo_raw_to_maxembed.py` | 18 | 🔴 |
+| 14 | L0 預設 16384 (64MB)，但 paper 聲稱 4096/16MB | `race_trace_v3.c` `L0_CAPACITY` | 8 | 🟡 |
+| 15 | Alibaba v1 converter（outfit-based）需 delete | `convert_alibaba_ifashion.py` | 18 | 🟡 |
+
+---
+
+## 🔴 Baseline Fairness（MaxEmbed 被削弱）
+
+Agent 11 發現 4 個參數不對齊 paper AE：
+
+| 參數 | Paper AE | 我方 | 影響 |
+|------|----------|------|------|
+| `-n` threads | 64 | 1 | 64x 少 concurrency |
+| `-b` batch_size | 32 (Criteo) | 16 | 減半 |
+| `-c` cache_ratio | 0.01-0.4 sweep | 0.0039（低於 paper range）| off-regime |
+| `rep_ratio` | sweep 5 個 | single | 選最差 rep_ratio？ |
+
+**行動**：補 anchor run at `-n 64 -b 32 -c 0.01`，讓 MaxEmbed 跑在它被調過的 regime
+
+---
+
+## 📚 Must-cite Competitors（Agent 4, 5, 6）
+
+| Paper | Venue | 重疊 |
+|-------|-------|------|
+| **CARINA** | SIGMOD'25 | **~70% overlap** on LFU filter for DLRM CXL（最嚴重）|
+| **Fleche** | EuroSys'22 | "probability-based filter policy" 已用 filter cache 術語 |
+| **RomeFS** | SoCC'24 | Dual-path CXL.mem + CXL.io（非 DLRM，但協議層已被做）|
+| **Hotline** | MLSys'23 | DLRM + DRAM/SSD tiering，直��� predecessor |
+| **FlexEMR** | EuroSys'24 | 3-tier DRAM/CXL/SSD recsys |
+| **UGACHE** | SOSP'23 | Near-optimal hotness policy for embeddings |
+| **Merlin HPS** | NVIDIA | 3-tier GPU/CPU/SSD 已建立 pattern |
+| **RecShard** | ASPLOS'22 | HBM+DRAM embedding sharding |
+| **RecSSD** | ASPLOS'21 | SSD embedding caching |
+| **TPP / Pond / TMO / HeMem** | ASPLOS'22-23, SOSP'21 | Generic OS tiering（orthogonal）|
+
+**我方剩下的 novelty**：**CXL-SSD × DLRM × stall-aware** 三維交集（single-device dual-path + recsys-specific + offline+online hybrid）
+
+---
+
+## 🟡 Math / Statistical Rigor（Agent 2, 13, 14）
+
+| Issue | Agent |
+|-------|:----:|
+| `dual_path_math_theory.md` 基本空 | 2 |
+| Threshold 推導 0.64 vs 0.761 矛盾，中間推導缺失 | 2 |
+| `ĥₚ`, `α`, `f̄`, effective_parallelism 都 undefined | 2 |
+| Independence assumption 被 PC(p) 破壞（coupling） | 2 |
+| n=5 對 CV>10% 不夠，CV 21% 要 n≥17，CV 50% 要 n≥95 | 13 |
+| E+L0 Optane ConfigD CV=49.6%（run3 spike 39.5ms）系統擾動 | 13 |
+| Warmup 沒 drop，B0 冷啟 3.4ms 主導 p99 | 14 |
+| 5x 的 500 樣本只有 ~100 獨立（trace replay）| 14 |
+| 1/100 sampling rate，p9999 無法取得 | 14 |
+
+---
+
+## 🟡 Simulation Fidelity（Agent 15, 16）
+
+| Issue | Status |
+|-------|:----:|
+| FEMU direct-mapped vs CMM-H 8-way SA | 🟡 config 一行可修 `buffer_way=3` |
+| FEMU FIFO vs CMM-H LRU + MRU | 🔴 FEMU 無 LRU 實作，需寫 `lru.c` |
+| FEMU MEM MISS (60μs) < NAND (68μs) inverted | 🔴 已知 bug，絕對數不可信 |
+| FEMU buffer 5MB vs CMM-H 48GB（1:10000）| 🟡 residency-budget framing 可辯護 |
+| SPDK TCP loopback 加 10-15μs/IO vs PCIe 5μs | 🟡 conservative 偏向（compressed speedup）|
+| SPDK p99 = avg（真 P4510 p99 ≈ 2.5x avg）| 🟡 對我方 claim 保守 |
+| SPDK `ch_xfer_lat = 0`（真 ONFI 3-4μs）| 🟡 minor |
+
+---
+
+## 📊 Verdict
+
+**Paper 目前狀態：不可直接送審**
+
+主要問題：
+1. **核心結論數字錯（Optane invert）**
+2. **Paper 捏造 stall + scheduling 實驗數字**
+3. **Narrative 描述的方法跟實際 code 不一致**
+4. **Tail latency 比較不合法**
+5. **Dataset 規模 2 個不對 paper**
+
+建議修復順序（優先級最高 → 最低）：
+1. 🚨 修 RESULTS.md Optane row + 重新算勝負表
+2. 🚨 重寫 paper_story.md / algorithm_design.md 為 "E+L0 + Host LFU L0"
+3. 🚨 刪 paper_story 捏造 stall 數字 + 改寫 scheduling table（或 rerun 補實驗）
+4. 🚨 改 pollution_model 為 "future work / analytic bound" 或實作
+5. 🔴 修 code bugs（Mode 2 timer, nhit=nm, LFU tie-break, encoding）
+6. 🔴 補 MaxEmbed `-n 64 -b 32 -c 0.01` anchor run
+7. 🔴 加 CARINA / Fleche / RomeFS / Hotline 等 related work
+8. 🟡 補 Criteo TB Tianchi full + Alibaba items 正確標註
+9. 🟡 修 FEMU buffer_way=3 + 考慮 LRU policy
+10. 🟡 補 tail latency 正確 instrumentation
+
+預估整體修復 **1-2 週**（主要 paper 重寫 + rerun scheduling 實驗 + bug fix）。
+
+---
+
+## 📋 20 Agents 各自 Report Links
+
+| Agent | Focus | Flag level |
+|-------|-------|:----------:|
+| 1 | Narrative audit | 🔴 |
+| 2 | Math rigor | 🔴 |
+| 3 | Claims vs data | 💀 |
+| 4 | CXL+DLRM prior art | 🔴 |
+| 5 | L0 filter prior art | 🔴 |
+| 6 | Dual-path prior art | 🔴 |
+| 7 | race_trace_v3.c audit | 🔴 |
+| 8 | LFU L0 impl | 🔴 |
+| 9 | FEMU mods | 🟡 |
+| 10 | Timer placement | 🟡 |
+| 11 | MaxEmbed baseline | 🔴 |
+| 12 | HW param P4510 | 🟡 |
+| 13 | Variance CV | 💀 |
+| 14 | Tail latency | 💀 |
+| 15 | CMM-H fidelity | 🔴 |
+| 16 | SPDK TCP overhead | 🟡 |
+| 17 | Dataset scale | 🔴 |
+| 18 | Preprocess alignment | 🔴 |
+| 19 | Claim ↔ code | 💀 |
+| 20 | Figure → data | 💀 |
+
+**💀 = catastrophic blocker**  
+**🔴 = significant flag**  
+**🟡 = concern, fixable**

METRICS_BACKLOG.md

@@ -0,0 +1,52 @@
+# Paper Metrics Backlog（待討論）
+
+以下為 paper 寫作可能需要補的 metrics，**尚未排入執行**。等當前 5 agents（#42/#43/#46/#47/#48）回來後再一起討論優先順序。
+
+## 1. Cache 替換算法 sweep
+- **Range**: LRU vs LFU vs CLOCK vs S3FIFO vs FIFO
+- **Scope**: E+L0 獨有的 L0 部分（host DRAM 16MB）
+- **Why**: 目前固定 LFU；reviewer 可能問「換成 LRU 就贏不了?」
+- **Tool**: race_trace_v3 有 `replacement` 參數支援多種（1=LIFO, 2=FIFO, 3=S3FIFO, 4=CLOCK）+ L0 是 LFU
+- **Expected runs**: 5 algo × 5 run = 25 runs per dataset × 5 datasets = 125 runs
+
+## 2. Replication ratio 詳細 sweep
+- **Range**: 0% / 5% / 10% / 20% / 80%（paper AE exp1 的 rep_ratios）
+- **Scope**: 每個 dataset 的 MaxEmbed 5x variance
+- **Note**: Task #47（exp1 matrix）已涵蓋（5 rep × 8 cache × 5 dataset = 200 runs），但這條獨立跑可 isolate rep ratio 效應
+- **Expected runs**: 5 rep × 5 run = 25 runs per dataset × 5 datasets = 125 runs
+
+## 3. Drift 跨 dataset
+- **Current**: CriteoTB d10-d20 (11 天 × 2 modes allmem/el0)
+- **Extend to**: Avazu drift (已有 `traces/avazu/avazu_drift/`)、Taobao drift (已有 `traces/taobao/taobao_drift/`)
+- **Why**: 證明 temporal drift 非 Criteo 獨有
+- **Expected runs**: 
+  - Avazu: 3 eval windows (day4-6, 7-8, 9-10) × 2 methods = 6 runs
+  - Taobao: 2 eval windows (day4-6, 7-9) × 2 methods = 4 runs
+  - = 10 runs
+
+## 4. Latency distribution (p50 / p99 / p999)
+- **Current**: 只有 mean per 5x variance
+- **Extend**: 從 per-batch CSV 算 p50 / p99 / p999 / p9999
+- **Why**: mean 隱藏 tail，reviewer 會問 tail behavior
+- **Tool**: 直接分析現有 `batch_latencies.csv`（race_trace_v3 per-batch 輸出）
+- **No new run needed** — 後處理現有資料
+- **Expected runs**: 0（純分析）
+
+---
+
+## 現在 running
+
+| # | Task | Agent ID |
+|---|------|----------|
+| #46 | Alibaba-iFashion POG→TSV converter | a6218ea79decef012 |
+| #42 | Avazu paper-scale processing | a55b66a42443b0cfe |
+| #43 | CriteoTB 22 days processing | a006ae63043d9cccf |
+| #48 | Thread sweep (-n) | a6097bcc3615f147d |
+| #47 | exp1 matrix (5×5×8=200 runs) | 未啟動，等 dataset 到位 |
+
+## 討論 checkpoint
+
+等上述 4 agents 都回來，我們再一起看：
+1. 每個 backlog metric 的 priority
+2. 能不能平行跑
+3. 要不要 FEMU multi-config（切 NAND / P4510 / Optane 的 cache algo sweep）

RESULTS.md

@@ -0,0 +1,111 @@
+# MaxEmbed vs E+L0 — 5x Variance Results (CORRECTED 2026-04-19)
+
+⚠️ **重要修正** (2026-04-19 audit)：Optane row 的 MaxEmbed 數字之前 parse 錯（unit swap）。重算後結論 **INVERT**：E+L0 在 Optane 也贏，不是輸。
+
+⚠️ **以下數據基於修復前 race_trace_v3.c + SPDK config，code bug 修完後要重跑**（見 `docs/experiments/rerun_required.md`）。
+
+所有實驗 5-run，每 run 前 hygiene（CPU governor=performance, drop_caches=3, numactl node 0）。
+
+---
+
+## 實驗數據（CORRECTED）
+
+| Config | NAND read lat | Method | Mean (overall_us) | CV | Per-sample | Ops/s | Raw CSV |
+|--------|---------------|--------|---|-----|------------|-------|---------|
+| Optane P5800X | 5μs | **MaxEmbed** | **15.80 ms/batch** | **21.3%** | 5.62 μs | 178,422 | `maxembed_release_ckfull_21mb_20260419_042951_run{1..5}.log` (CSV empty — data in logs only) |
+| Optane P5800X | 5μs | E+L0 | 1.66 ms/batch | 49.6% ⚠️ | 12.97 μs | ~77,000 | `eplus_l0_configD_optane_20260419_041958_fixed.csv` (main CSV corrupt — "Avg:" strings) |
+| TLC NAND cell | 40μs | MaxEmbed | 4.77 ms/batch | 2.2% | 298 μs | 3,368 | `maxembed_nand_21mb_20260419_045612.csv` |
+| TLC NAND cell | 40μs | E+L0 | 0.92 ms/batch | 4.9% | 57.5 μs | ~17,400 | `eplus_l0_nand_21mb_v2_20260419_051219.csv` |
+| **P4510 (paper HW)** | **85μs** | **MaxEmbed** | **5.22 ms/batch** | **1.4%** | **326 μs** | **3,079** | `maxembed_p4510_21mb_20260419_165216.csv` |
+| **P4510 (paper HW)** | **85μs** | **E+L0** | **1.20 ms/batch** | **~0%** | **75 μs** | **~13,300** | `eplus_l0_p4510_21mb_20260419_165558.csv` |
+
+---
+
+## 勝負表（CORRECTED）
+
+| Hardware | MaxEmbed ms/batch | E+L0 ms/batch | 勝者 | 倍數 |
+|----------|:-----------------:|:--------------:|:----:|:----:|
+| **Optane 5μs** | **15.80** | **1.66** | **E+L0** | **9.52x** ⭐ |
+| NAND 40μs | 4.77 | 0.92 | **E+L0** | 5.18x |
+| P4510 85μs (paper HW) | 5.22 | 1.20 | **E+L0** | 4.35x |
+
+**結論**：E+L0 在**全部 3 個 HW config** 都贏 MaxEmbed。Optane 修正前錯標為 "MaxEmbed 贏 2.31x"，實際是 **E+L0 贏 9.52x**。
+
+---
+
+## 已知 caveats（20-agent audit 結果）
+
+### 💀 Critical
+1. **code bug 未修**：race_trace_v3.c Mode 2 有 timer 包 malloc + `nhit=nm` 假 miss + lfence asymmetry，修完後全部 rerun
+2. **CV 21.3%（Optane MaxEmbed）和 49.6%（E+L0 Optane ConfigD）過高**：n=5 不夠統計顯著，要 n≥17（Optane）/ n≥95（ConfigD）
+3. **E+L0 Optane ConfigD 有 run3/run5 mid-trace spike（39.5ms / 19.6ms）**：需調查系統擾動 vs algorithm
+
+### 🔴 Major
+4. **MaxEmbed baseline 4 參數削弱**（`-n 1` / `-b 16` / `-c 0.0039` / single rep_ratio），要補 anchor run 對齊 paper AE
+5. **MaxEmbed per-batch tail 不可比**：MaxEmbed 報 100-200 batch windowed avg，E+L0 報 true per-batch
+6. **`/tmp/batch_latencies.csv` 沒存**，需 instrument per-batch CSV
+
+### 🟡 Minor
+7. SPDK TCP loopback 加 10-15μs/IO（vs PCIe 5μs）：對結論保守
+8. SPDK p99=avg 不真實（真 P4510 p99 ~2.5x avg）
+9. FEMU direct-mapped vs CMM-H 8-way（config `buffer_way=3` 一行可修）
+
+---
+
+## Config 細節
+
+**Workload**: CK-Full per-sample queries（Criteo Kaggle 25.67M samples, 26 features/sample, 22M unique items, batch=16）
+
+**Shared**:
+- Embedding dim = 64 (float32, 256 bytes/entry)
+- Cache budget = 21 MB（85,939 entries = MaxEmbed CacheLib）
+- Single thread（`-n 1`）← 削弱 MaxEmbed，要補 `-n 64` anchor
+
+**MaxEmbed**：
+- `--cache_ratio 0.0039` ← 削弱 MaxEmbed，paper AE 用 0.01-0.4 sweep
+- Transport: SPDK NVMe-oF over TCP loopback
+- Backend: bdev_delay → bdev_null（avg_read = pg_rd_lat, avg_write = pg_wr_lat）
+- Model delay: `--delay 150`（MLP busy-wait 150μs/batch）
+- Build: Release + `-DNDEBUG`（`x64-linux-release` triplet）
+
+**E+L0**（FEMU CXL-SSD Mode 2）：
+- L0 = 4096 host pages (16MB LFU)，Device DRAM buffer = 5MB
+- K=12, R=4（hot threshold）
+- `io_threads=8`（effective QD）
+- MLP delay argv[9]=150
+- Remap: `ck_se_remap.json`（SeedExpand）
+
+**FEMU**：
+- `femu_mode=6` (CXL-SSD)
+- 8ch × 8lun × 1pl × 96blk × 256pg × 8sec × 512B ≈ 6GB
+- `bufsz_mb=5`, `replacement=2` (FIFO) ← 要改 8-way LRU 對齊 CMM-H
+- `prefetch_degree=0`
+- `pg_rd_lat` / `pg_wr_lat` per config
+
+---
+
+## 實驗數據如何記錄（recording methodology）
+
+### 1. 原始資料（timestamped raw logs）
+**Path**：`research_data/results/criteo_kaggle/variance_5x_host/{desc}_{YYYYMMDD_HHMMSS}_run{1..5}.log`
+
+每個 run 的完整 stdout／stderr（包含 per-batch breakdown、final Result 行）。不覆蓋，永久存檔。
+
+### 2. Parsed CSV
+**Path**：同上 `{desc}_{YYYYMMDD_HHMMSS}.csv`
+
+Schema（**unit 要明確避免再發生 parse error**）：
+- MaxEmbed: `run, overall_us (batch latency μs), ops_per_sec, prepare_us, calc_us, ssd_us, model_us`
+- E+L0: `run, avg_ms_per_batch, mem_hit_rate, mem_total, io_total, l0_total, batches`
+
+**注意**：`overall_us` 就是 batch latency 毫秒值，不是 per-sample。per-sample = `1e6 / ops_per_sec`。
+
+### 3. Runner 腳本
+`research_data/scripts/maxembed/run_{maxembed,el0}_5x.sh`
+
+### 4. 本檔（彙總）
+`research_data/results/RESULTS.md`（本檔）— canonical table，每次 rerun 更新
+
+### 5. HW config 檔
+- FEMU：`research_data/configs/femu/run-cxlssd-{optane,prod,p4510}.sh`
+- SPDK：執行期 RPC，config 存 daemon memory，要 snapshot 到 `research_data/configs/spdk/history.md`

beta_fine_sweep/run_20260501_012552.log

@@ -0,0 +1,198 @@
+[01:25:52] === START β fine sweep ===
+[01:25:52] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[01:25:52] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[01:27:09] [criteo_kaggle] β=0.50 → K=21 R=1
+[01:28:07] [criteo_kaggle] β=0.55 → K=19 R=1
+[01:29:05] [criteo_kaggle] β=0.60 → K=17 R=1
+[01:30:02] [criteo_kaggle] β=0.65 → K=16 R=1
+[01:31:04] [criteo_kaggle] β=0.70 → K=15 R=1
+[01:32:08] [criteo_kaggle] β=0.7311 → K=14 R=1
+[01:33:14] [criteo_kaggle] β=0.75 → K=14 R=1
+[01:34:21] [criteo_kaggle] β=0.80 → K=13 R=1
+[01:35:28] [criteo_kaggle] β=0.85 → K=12 R=1
+[01:36:37] [criteo_kaggle] β=0.90 → K=12 R=1
+[01:37:47] [criteo_kaggle] === DONE ===
+[01:37:47] [avazu] ws=2307MB bufsz=108 l0=2764
+[01:37:47] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[01:39:13] [avazu] β=0.50 → K=298 R=1
+[01:39:21] [avazu] β=0.55 → K=296 R=1
+[01:39:28] [avazu] β=0.60 → K=293 R=1
+[01:39:36] [avazu] β=0.65 → K=291 R=1
+[01:39:43] [avazu] β=0.70 → K=287 R=1
+[01:39:51] [avazu] β=0.7311 → K=283 R=1
+[01:39:58] [avazu] β=0.75 → K=275 R=1
+[01:40:06] [avazu] β=0.80 → K=261 R=1
+[01:40:14] [avazu] β=0.85 → K=260 R=1
+[01:40:21] [avazu] β=0.90 → K=259 R=1
+[01:40:28] [avazu] === DONE ===
+[01:40:28] [amazon_m2] ws=341MB bufsz=15 l0=384
+[01:40:28] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[01:41:54] [amazon_m2] β=0.50 → K=8 R=9
+[01:42:01] [amazon_m2] β=0.55 → K=8 R=8
+[01:42:08] [amazon_m2] β=0.60 → K=8 R=7
+[01:42:14] [amazon_m2] β=0.65 → K=8 R=6
+[01:42:21] [amazon_m2] β=0.70 → K=8 R=6
+[01:42:28] [amazon_m2] β=0.7311 → K=7 R=5
+[01:42:34] [amazon_m2] β=0.75 → K=7 R=5
+[01:42:41] [amazon_m2] β=0.80 → K=7 R=4
+[01:42:48] [amazon_m2] β=0.85 → K=7 R=3
+[01:42:54] [amazon_m2] β=0.90 → K=7 R=2
+[01:43:01] [amazon_m2] === DONE ===
+[01:43:01] [alibaba_ifashion] ws=532MB bufsz=24 l0=614
+[01:43:01] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[01:44:27] [alibaba_ifashion] β=0.50 → K=10 R=1
+[01:45:05] [alibaba_ifashion] β=0.55 → K=9 R=1
+[01:45:43] [alibaba_ifashion] β=0.60 → K=9 R=1
+[01:46:22] [alibaba_ifashion] β=0.65 → K=9 R=1
+[01:47:02] [alibaba_ifashion] β=0.70 → K=9 R=1
+[01:47:41] [alibaba_ifashion] β=0.7311 → K=9 R=1
+[01:48:20] [alibaba_ifashion] β=0.75 → K=8 R=1
+[01:49:03] [alibaba_ifashion] β=0.80 → K=8 R=1
+[01:49:45] [alibaba_ifashion] β=0.85 → K=8 R=1
+[01:50:27] [alibaba_ifashion] β=0.90 → K=8 R=1
+[01:51:09] [alibaba_ifashion] === DONE ===
+[01:51:09] [tmall] ws=531MB bufsz=24 l0=614
+[01:51:09] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[01:52:31] [tmall] β=0.50 → K=14 R=1
+[01:52:41] [tmall] β=0.55 → K=14 R=1
+[01:52:51] [tmall] β=0.60 → K=13 R=1
+[01:53:01] [tmall] β=0.65 → K=13 R=1
+[01:53:11] [tmall] β=0.70 → K=13 R=1
+[01:53:22] [tmall] β=0.7311 → K=12 R=1
+[01:53:32] [tmall] β=0.75 → K=12 R=1
+[01:53:42] [tmall] β=0.80 → K=12 R=1
+[01:53:51] [tmall] β=0.85 → K=12 R=1
+[01:54:02] [tmall] β=0.90 → K=11 R=1
+[01:54:12] [tmall] === DONE ===
+[01:54:12] [gowalla] ws=313MB bufsz=14 l0=358
+[01:54:12] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[01:55:38] [gowalla] β=0.50 → K=5 R=1
+[01:55:49] [gowalla] β=0.55 → K=5 R=1
+[01:55:59] [gowalla] β=0.60 → K=5 R=1
+[01:56:10] [gowalla] β=0.65 → K=5 R=1
+[01:56:20] [gowalla] β=0.70 → K=5 R=1
+[01:56:31] [gowalla] β=0.7311 → K=5 R=1
+[01:56:41] [gowalla] β=0.75 → K=5 R=1
+[01:56:51] [gowalla] β=0.80 → K=4 R=1
+[01:57:01] [gowalla] β=0.85 → K=4 R=1
+[01:57:11] [gowalla] β=0.90 → K=4 R=1
+[01:57:21] [gowalla] === DONE ===
+[01:57:21] [lastfm_1k] ws=309MB bufsz=14 l0=358
+[01:57:21] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[01:58:48] [lastfm_1k] β=0.50 → K=31 R=1
+[01:59:02] [lastfm_1k] β=0.55 → K=29 R=1
+[01:59:15] [lastfm_1k] β=0.60 → K=27 R=1
+[01:59:29] [lastfm_1k] β=0.65 → K=25 R=1
+[01:59:43] [lastfm_1k] β=0.70 → K=23 R=1
+[01:59:57] [lastfm_1k] β=0.7311 → K=22 R=1
+[02:00:10] [lastfm_1k] β=0.75 → K=22 R=1
+[02:00:24] [lastfm_1k] β=0.80 → K=21 R=1
+[02:00:37] [lastfm_1k] β=0.85 → K=19 R=1
+[02:00:51] [lastfm_1k] β=0.90 → K=18 R=1
+[02:01:04] [lastfm_1k] === DONE ===
+[02:01:04] [taobao] ws=102MB bufsz=4 l0=102
+[02:01:04] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[02:02:25] [taobao] β=0.50 → K=38 R=1
+[02:02:33] [taobao] β=0.55 → K=37 R=1
+[02:02:40] [taobao] β=0.60 → K=36 R=1
+[02:02:48] [taobao] β=0.65 → K=35 R=1
+[02:02:55] [taobao] β=0.70 → K=34 R=1
+[02:03:03] [taobao] β=0.7311 → K=33 R=1
+[02:03:10] [taobao] β=0.75 → K=33 R=1
+[02:03:18] [taobao] β=0.80 → K=32 R=1
+[02:03:26] [taobao] β=0.85 → K=31 R=1
+[02:03:33] [taobao] β=0.90 → K=30 R=1
+[02:03:40] [taobao] === DONE ===
+[02:03:40] [bookcrossing] ws=81MB bufsz=3 l0=76
+[02:03:40] [bookcrossing] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[02:05:02] [bookcrossing] β=0.50 → K=13 R=1
+[02:05:07] [bookcrossing] β=0.55 → K=12 R=1
+[02:05:12] [bookcrossing] β=0.60 → K=11 R=1
+[02:05:16] [bookcrossing] β=0.65 → K=11 R=1
+[02:05:21] [bookcrossing] β=0.70 → K=10 R=1
+[02:05:26] [bookcrossing] β=0.7311 → K=10 R=1
+[02:05:31] [bookcrossing] β=0.75 → K=10 R=1
+[02:05:37] [bookcrossing] β=0.80 → K=9 R=1
+[02:05:42] [bookcrossing] β=0.85 → K=9 R=1
+[02:05:47] [bookcrossing] β=0.90 → K=9 R=1
+[02:05:52] [bookcrossing] === DONE ===
+[02:05:52] [lastfm_360k] ws=67MB bufsz=3 l0=76
+[02:05:52] [lastfm_360k] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[02:07:18] [lastfm_360k] β=0.50 → K=9 R=1
+[02:07:28] [lastfm_360k] β=0.55 → K=9 R=1
+[02:07:37] [lastfm_360k] β=0.60 → K=9 R=1
+[02:07:47] [lastfm_360k] β=0.65 → K=8 R=1
+[02:07:56] [lastfm_360k] β=0.70 → K=8 R=1
+[02:08:05] [lastfm_360k] β=0.7311 → K=8 R=1
+[02:08:15] [lastfm_360k] β=0.75 → K=8 R=1
+[02:08:24] [lastfm_360k] β=0.80 → K=8 R=1
+[02:08:33] [lastfm_360k] β=0.85 → K=8 R=1
+[02:08:42] [lastfm_360k] β=0.90 → K=8 R=1
+[02:08:51] [lastfm_360k] === DONE ===
+[02:08:51] [amazon_Electronics] ws=63MB bufsz=2 l0=51
+[02:08:51] [amazon_Electronics] reboot VM ssd=6144 bufsz=2MB
+/dev/dax0.0
+[02:10:17] [amazon_Electronics] β=0.50 → K=14 R=1
+[02:10:22] [amazon_Electronics] β=0.55 → K=13 R=1
+[02:10:27] [amazon_Electronics] β=0.60 → K=13 R=1
+[02:10:32] [amazon_Electronics] β=0.65 → K=13 R=1
+[02:10:37] [amazon_Electronics] β=0.70 → K=13 R=1
+[02:10:42] [amazon_Electronics] β=0.7311 → K=12 R=1
+[02:10:47] [amazon_Electronics] β=0.75 → K=12 R=1
+[02:10:51] [amazon_Electronics] β=0.80 → K=12 R=1
+[02:10:56] [amazon_Electronics] β=0.85 → K=11 R=1
+[02:11:00] [amazon_Electronics] β=0.90 → K=11 R=1
+[02:11:05] [amazon_Electronics] === DONE ===
+[02:11:05] [yelp] ws=37MB bufsz=1 l0=25
+[02:11:05] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[02:12:26] [yelp] β=0.50 → K=6 R=1
+[02:12:31] [yelp] β=0.55 → K=6 R=1
+[02:12:36] [yelp] β=0.60 → K=6 R=1
+[02:12:41] [yelp] β=0.65 → K=6 R=1
+[02:12:45] [yelp] β=0.70 → K=5 R=1
+[02:12:49] [yelp] β=0.7311 → K=5 R=1
+[02:12:53] [yelp] β=0.75 → K=5 R=1
+[02:12:58] [yelp] β=0.80 → K=5 R=1
+[02:13:03] [yelp] β=0.85 → K=5 R=1
+[02:13:07] [yelp] β=0.90 → K=5 R=1
+[02:13:11] [yelp] === DONE ===
+[02:13:11] [retailrocket] ws=36MB bufsz=1 l0=25
+[02:13:11] [retailrocket] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[02:14:33] [retailrocket] β=0.50 → K=14 R=2
+[02:14:39] [retailrocket] β=0.55 → K=13 R=2
+[02:14:44] [retailrocket] β=0.60 → K=13 R=2
+[02:14:50] [retailrocket] β=0.65 → K=13 R=1
+[02:14:56] [retailrocket] β=0.70 → K=12 R=1
+[02:15:02] [retailrocket] β=0.7311 → K=12 R=1
+[02:15:08] [retailrocket] β=0.75 → K=12 R=1
+[02:15:14] [retailrocket] β=0.80 → K=12 R=1
+[02:15:20] [retailrocket] β=0.85 → K=12 R=1
+[02:15:26] [retailrocket] β=0.90 → K=11 R=1
+[02:15:31] [retailrocket] === DONE ===
+[02:15:31] [foursquare] ws=25MB bufsz=1 l0=25
+[02:15:31] [foursquare] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[02:17:03] [foursquare] β=0.50 → K=11 R=1
+[02:17:14] [foursquare] β=0.55 → K=10 R=1
+[02:17:25] [foursquare] β=0.60 → K=10 R=1
+[02:17:35] [foursquare] β=0.65 → K=10 R=1
+[02:17:46] [foursquare] β=0.70 → K=10 R=1
+[02:17:57] [foursquare] β=0.7311 → K=9 R=1
+[02:18:08] [foursquare] β=0.75 → K=9 R=1
+[02:18:18] [foursquare] β=0.80 → K=9 R=1
+[02:18:29] [foursquare] β=0.85 → K=9 R=1
+[02:18:39] [foursquare] β=0.90 → K=9 R=1
+[02:18:50] [foursquare] === DONE ===
+[02:18:50] === ALL DONE ===

criteo_tb_d0_10.log

@@ -0,0 +1,193 @@
+[23:58:41] === process days 0-10 ===
+[skip] day_0 already processed
+[23:58:41] processing day_1 (16G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d1
+Criteo Terabyte → Embedding Trace
+  Days: [1]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d1
+  Reading day_1.gz...
+    day_1: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 219,729 features
+    Table 9: 164,599 features
+    Table 19: 230,453 features
+    Table 20: 95,176 features
+    Table 21: 201,182 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d1/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d1
+[23:59:13] day_1 done: 102M
+[23:59:13] processing day_2 (16G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d2
+Criteo Terabyte → Embedding Trace
+  Days: [2]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d2
+  Reading day_2.gz...
+    day_2: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 201,176 features
+    Table 9: 153,770 features
+    Table 19: 210,963 features
+    Table 20: 93,020 features
+    Table 21: 184,497 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d2/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d2
+[23:59:49] day_2 done: 101M
+[23:59:49] processing day_3 (15G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d3
+Criteo Terabyte → Embedding Trace
+  Days: [3]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d3
+  Reading day_3.gz...
+    day_3: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 204,027 features
+    Table 9: 155,372 features
+    Table 19: 213,965 features
+    Table 20: 92,989 features
+    Table 21: 187,037 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d3/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d3
+[00:00:21] day_3 done: 103M
+[00:00:21] processing day_4 (12G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d4
+Criteo Terabyte → Embedding Trace
+  Days: [4]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d4
+  Reading day_4.gz...
+    day_4: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 241,791 features
+    Table 9: 180,337 features
+    Table 19: 253,893 features
+    Table 20: 104,516 features
+    Table 21: 220,420 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d4/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d4
+[00:00:51] day_4 done: 104M
+[00:00:51] processing day_5 (14G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d5
+Criteo Terabyte → Embedding Trace
+  Days: [5]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d5
+  Reading day_5.gz...
+    day_5: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 246,935 features
+    Table 9: 186,045 features
+    Table 19: 259,191 features
+    Table 20: 106,398 features
+    Table 21: 225,625 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d5/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d5
+[00:01:27] day_5 done: 104M
+[00:01:27] processing day_6 (16G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d6
+Criteo Terabyte → Embedding Trace
+  Days: [6]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d6
+  Reading day_6.gz...
+    day_6: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 223,983 features
+    Table 9: 168,384 features
+    Table 19: 234,589 features
+    Table 20: 97,775 features
+    Table 21: 205,193 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d6/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d6
+[00:02:00] day_6 done: 100M
+[00:02:00] processing day_7 (16G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d7
+Criteo Terabyte → Embedding Trace
+  Days: [7]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d7
+  Reading day_7.gz...
+    day_7: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 215,648 features
+    Table 9: 164,976 features
+    Table 19: 226,015 features
+    Table 20: 98,785 features
+    Table 21: 198,496 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d7/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d7
+[00:02:36] day_7 done: 102M
+[00:02:36] processing day_8 (16G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d8
+Criteo Terabyte → Embedding Trace
+  Days: [8]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d8
+  Reading day_8.gz...
+    day_8: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 216,790 features
+    Table 9: 165,348 features
+    Table 19: 227,754 features
+    Table 20: 98,737 features
+    Table 21: 199,325 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d8/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d8
+[00:03:09] day_8 done: 103M
+[00:03:10] processing day_9 (16G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d9
+Criteo Terabyte → Embedding Trace
+  Days: [9]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d9
+  Reading day_9.gz...
+    day_9: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 230,752 features
+    Table 9: 172,125 features
+    Table 19: 242,481 features
+    Table 20: 100,309 features
+    Table 21: 210,995 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d9/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d9
+[00:03:46] day_9 done: 101M
+[00:03:46] processing day_10 (15G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d10
+Criteo Terabyte → Embedding Trace
+  Days: [10]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d10
+  Reading day_10.gz...
+    day_10: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 230,246 features
+    Table 9: 171,870 features
+    Table 19: 242,295 features
+    Table 20: 99,443 features
+    Table 21: 210,520 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d10/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d10
+[00:04:19] day_10 done: 103M
+[00:04:19] days 0-10 done

criteo_tb_watcher.log

@@ -0,0 +1,41 @@
+[23:58:41] === START watcher (days 21,22,23) ===
+[00:35:42] day_23.gz stable, processing...
+[00:35:42] processing day_23 (14G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d23
+Criteo Terabyte → Embedding Trace
+  Days: [23]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d23
+  Reading day_23.gz...
+    day_23: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 232,072 features
+    Table 9: 173,481 features
+    Table 19: 244,680 features
+    Table 20: 102,962 features
+    Table 21: 210,696 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d23/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d23
+[00:36:12] day_23 done: 104M
+[00:46:42] day_22.gz stable, processing...
+[00:46:42] processing day_22 (15G) → /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d22
+Criteo Terabyte → Embedding Trace
+  Days: [22]
+  Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d22
+  Reading day_22.gz...
+    day_22: 1,280,000 rows
+
+  Total: 1,280,000 records across 1 days
+    Table 0: 236,227 features
+    Table 9: 176,555 features
+    Table 19: 248,507 features
+    Table 20: 103,041 features
+    Table 21: 214,284 features
+
+  Generating inference trace (10000 batches)...
+  /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d22/inference_trace.csv: 33,280,000 accesses
+
+Done. Output: /home/sadoo/Projects/kvcache/research_data/traces/criteo_terabyte/criteo_terabyte_d22
+[00:47:13] day_22 done: 102M

ewma_hyperparam/run_20260501_021850.log

@@ -0,0 +1,89 @@
+[02:18:50] === START EWMA hyperparam sweep ===
+[02:18:50] [taobao] ws=102MB bufsz=4 l0=102
+[02:18:50] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[02:20:16] [taobao] α=0.05 warmup=0
+[02:20:19] [taobao] α=0.05 warmup=50
+[02:20:22] [taobao] α=0.05 warmup=100
+[02:20:25] [taobao] α=0.05 warmup=200
+[02:20:27] [taobao] α=0.05 warmup=500
+[02:20:30] [taobao] α=0.10 warmup=0
+[02:20:33] [taobao] α=0.10 warmup=50
+[02:20:35] [taobao] α=0.10 warmup=100
+[02:20:38] [taobao] α=0.10 warmup=200
+[02:20:41] [taobao] α=0.10 warmup=500
+[02:20:44] [taobao] α=0.20 warmup=0
+[02:20:47] [taobao] α=0.20 warmup=50
+[02:20:50] [taobao] α=0.20 warmup=100
+[02:20:53] [taobao] α=0.20 warmup=200
+[02:20:56] [taobao] α=0.20 warmup=500
+[02:20:59] [taobao] α=0.30 warmup=0
+[02:21:01] [taobao] α=0.30 warmup=50
+[02:21:05] [taobao] α=0.30 warmup=100
+[02:21:07] [taobao] α=0.30 warmup=200
+[02:21:10] [taobao] α=0.30 warmup=500
+[02:21:13] [taobao] α=0.50 warmup=0
+[02:21:16] [taobao] α=0.50 warmup=50
+[02:21:18] [taobao] α=0.50 warmup=100
+[02:21:21] [taobao] α=0.50 warmup=200
+[02:21:24] [taobao] α=0.50 warmup=500
+[02:21:27] [taobao] === DONE ===
+[02:21:27] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[02:21:27] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[02:22:53] [criteo_kaggle] α=0.05 warmup=0
+[02:23:15] [criteo_kaggle] α=0.05 warmup=50
+[02:23:36] [criteo_kaggle] α=0.05 warmup=100
+[02:23:57] [criteo_kaggle] α=0.05 warmup=200
+[02:24:19] [criteo_kaggle] α=0.05 warmup=500
+[02:24:43] [criteo_kaggle] α=0.10 warmup=0
+[02:25:03] [criteo_kaggle] α=0.10 warmup=50
+[02:25:23] [criteo_kaggle] α=0.10 warmup=100
+[02:25:45] [criteo_kaggle] α=0.10 warmup=200
+[02:26:06] [criteo_kaggle] α=0.10 warmup=500
+[02:26:30] [criteo_kaggle] α=0.20 warmup=0
+[02:26:51] [criteo_kaggle] α=0.20 warmup=50
+[02:27:13] [criteo_kaggle] α=0.20 warmup=100
+[02:27:33] [criteo_kaggle] α=0.20 warmup=200
+[02:27:54] [criteo_kaggle] α=0.20 warmup=500
+[02:28:19] [criteo_kaggle] α=0.30 warmup=0
+[02:28:39] [criteo_kaggle] α=0.30 warmup=50
+[02:29:00] [criteo_kaggle] α=0.30 warmup=100
+[02:29:22] [criteo_kaggle] α=0.30 warmup=200
+[02:29:43] [criteo_kaggle] α=0.30 warmup=500
+[02:30:07] [criteo_kaggle] α=0.50 warmup=0
+[02:30:27] [criteo_kaggle] α=0.50 warmup=50
+[02:30:48] [criteo_kaggle] α=0.50 warmup=100
+[02:31:12] [criteo_kaggle] α=0.50 warmup=200
+[02:31:34] [criteo_kaggle] α=0.50 warmup=500
+[02:31:57] [criteo_kaggle] === DONE ===
+[02:31:57] [avazu] ws=2307MB bufsz=108 l0=2764
+[02:31:57] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[02:33:23] [avazu] α=0.05 warmup=0
+[02:33:27] [avazu] α=0.05 warmup=50
+[02:33:30] [avazu] α=0.05 warmup=100
+[02:33:34] [avazu] α=0.05 warmup=200
+[02:33:38] [avazu] α=0.05 warmup=500
+[02:33:41] [avazu] α=0.10 warmup=0
+[02:33:45] [avazu] α=0.10 warmup=50
+[02:33:48] [avazu] α=0.10 warmup=100
+[02:33:52] [avazu] α=0.10 warmup=200
+[02:33:55] [avazu] α=0.10 warmup=500
+[02:33:59] [avazu] α=0.20 warmup=0
+[02:34:02] [avazu] α=0.20 warmup=50
+[02:34:06] [avazu] α=0.20 warmup=100
+[02:34:10] [avazu] α=0.20 warmup=200
+[02:34:14] [avazu] α=0.20 warmup=500
+[02:34:17] [avazu] α=0.30 warmup=0
+[02:34:21] [avazu] α=0.30 warmup=50
+[02:34:24] [avazu] α=0.30 warmup=100
+[02:34:28] [avazu] α=0.30 warmup=200
+[02:34:31] [avazu] α=0.30 warmup=500
+[02:34:35] [avazu] α=0.50 warmup=0
+[02:34:38] [avazu] α=0.50 warmup=50
+[02:34:42] [avazu] α=0.50 warmup=100
+[02:34:45] [avazu] α=0.50 warmup=200
+[02:34:49] [avazu] α=0.50 warmup=500
+[02:34:52] [avazu] === DONE ===
+[02:34:52] === ALL DONE ===

ewma_quant/run_20260430_183533.log

@@ -0,0 +1,23 @@
+[18:35:34] === START EWMA Quantization Ablation ===
+[18:35:34] [taobao] ws=102MB bufsz=4 l0=102
+[18:35:34] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[18:37:00] [taobao] mode=9
+[18:37:18] [taobao] mode=11
+[18:37:25] [taobao] mode=12
+[18:37:31] [taobao] === DONE ===
+[18:37:31] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[18:37:31] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[18:38:57] [criteo_kaggle] mode=9
+[18:39:34] [criteo_kaggle] mode=11
+[18:40:09] [criteo_kaggle] mode=12
+[18:40:42] [criteo_kaggle] === DONE ===
+[18:40:42] [avazu] ws=2307MB bufsz=108 l0=2764
+[18:40:42] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[18:42:09] [avazu] mode=9
+[18:42:15] [avazu] mode=11
+[18:42:21] [avazu] mode=12
+[18:42:27] [avazu] === DONE ===
+[18:42:27] === ALL DONE ===

ewma_quant_reverse/run_20260501_011345.log

@@ -0,0 +1,23 @@
+[01:13:45] === START EWMA Quant REVERSE-ORDER ===
+[01:13:45] [taobao] ws=102MB bufsz=4 l0=102
+[01:13:45] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[01:15:12] [taobao] mode=11
+[01:15:24] [taobao] mode=12
+[01:15:30] [taobao] mode=9
+[01:15:36] [taobao] === DONE ===
+[01:15:36] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[01:15:36] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[01:17:02] [criteo_kaggle] mode=11
+[01:17:39] [criteo_kaggle] mode=12
+[01:18:12] [criteo_kaggle] mode=9
+[01:18:47] [criteo_kaggle] === DONE ===
+[01:18:47] [avazu] ws=2307MB bufsz=108 l0=2764
+[01:18:47] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[01:20:13] [avazu] mode=11
+[01:20:18] [avazu] mode=12
+[01:20:22] [avazu] mode=9
+[01:20:26] [avazu] === DONE ===
+[01:20:26] === ALL DONE ===

f1_ablation/run_20260501_012052.log

@@ -0,0 +1,20 @@
+[01:20:52] === START F1 Ablation (Mode 9 vs 13) ===
+[01:20:52] [taobao] ws=102MB bufsz=4 l0=102
+[01:20:52] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[01:22:18] [taobao] mode=9
+[01:22:24] [taobao] mode=13
+[01:22:30] [taobao] === DONE ===
+[01:22:30] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[01:22:30] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[01:23:51] [criteo_kaggle] mode=9
+[01:24:03] [criteo_kaggle] mode=13
+[01:24:14] [criteo_kaggle] === DONE ===
+[01:24:14] [avazu] ws=2307MB bufsz=108 l0=2764
+[01:24:14] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[01:25:40] [avazu] mode=9
+[01:25:46] [avazu] mode=13
+[01:25:52] [avazu] === DONE ===
+[01:25:52] === ALL DONE ===

l0_sweep/run_20260501_031141.log

@@ -0,0 +1,83 @@
+[03:11:41] === START L0 size sweep ===
+[03:11:41] [criteo_kaggle] ws=5380MB bufsz=252 K=21 R=1
+[03:11:41] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[03:13:07] [criteo_kaggle] L0=5% (3225 pages) mode=2
+[03:13:19] [criteo_kaggle] L0=10% (6451 pages) mode=2
+[03:13:30] [criteo_kaggle] L0=20% (12902 pages) mode=2
+[03:13:41] [criteo_kaggle] L0=50% (32256 pages) mode=2
+[03:13:52] [criteo_kaggle] L0=100% (64512 pages) mode=2
+[03:14:04] [criteo_kaggle] === DONE ===
+[03:14:04] [avazu] ws=2307MB bufsz=108 K=298 R=1
+[03:14:04] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[03:15:30] [avazu] L0=5% (1382 pages) mode=2
+[03:15:39] [avazu] L0=10% (2764 pages) mode=2
+[03:15:46] [avazu] L0=20% (5529 pages) mode=2
+[03:15:54] [avazu] L0=50% (13824 pages) mode=2
+[03:16:00] [avazu] L0=100% (27648 pages) mode=2
+[03:16:07] [avazu] === DONE ===
+[03:16:07] [taobao] ws=102MB bufsz=4 K=38 R=1
+[03:16:07] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[03:17:33] [taobao] L0=5% (51 pages) mode=2
+[03:17:40] [taobao] L0=10% (102 pages) mode=2
+[03:17:48] [taobao] L0=20% (204 pages) mode=2
+[03:17:55] [taobao] L0=50% (512 pages) mode=2
+[03:18:02] [taobao] L0=100% (1024 pages) mode=2
+[03:18:10] [taobao] === DONE ===
+[03:18:10] [alibaba_ifashion] ws=532MB bufsz=24 K=10 R=1
+[03:18:10] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:19:26] [alibaba_ifashion] L0=5% (307 pages) mode=2
+[03:20:03] [alibaba_ifashion] L0=10% (614 pages) mode=2
+[03:20:41] [alibaba_ifashion] L0=20% (1228 pages) mode=2
+[03:21:18] [alibaba_ifashion] L0=50% (3072 pages) mode=2
+[03:21:55] [alibaba_ifashion] L0=100% (6144 pages) mode=2
+[03:22:30] [alibaba_ifashion] === DONE ===
+[03:22:30] [tmall] ws=531MB bufsz=24 K=14 R=1
+[03:22:30] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:23:56] [tmall] L0=5% (307 pages) mode=2
+[03:24:07] [tmall] L0=10% (614 pages) mode=2
+[03:24:17] [tmall] L0=20% (1228 pages) mode=2
+[03:24:27] [tmall] L0=50% (3072 pages) mode=2
+[03:24:36] [tmall] L0=100% (6144 pages) mode=2
+[03:24:46] [tmall] === DONE ===
+[03:24:46] [lastfm_1k] ws=309MB bufsz=14 K=31 R=1
+[03:24:46] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[03:26:02] [lastfm_1k] L0=5% (179 pages) mode=2
+[03:26:17] [lastfm_1k] L0=10% (358 pages) mode=2
+[03:26:30] [lastfm_1k] L0=20% (716 pages) mode=2
+[03:26:45] [lastfm_1k] L0=50% (1792 pages) mode=2
+[03:26:59] [lastfm_1k] L0=100% (3584 pages) mode=2
+[03:27:14] [lastfm_1k] === DONE ===
+[03:27:14] [yelp] ws=37MB bufsz=1 K=5 R=1
+[03:27:14] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[03:28:35] [yelp] L0=5% (12 pages) mode=2
+[03:28:44] [yelp] L0=10% (25 pages) mode=2
+[03:28:52] [yelp] L0=20% (51 pages) mode=2
+[03:29:01] [yelp] L0=50% (128 pages) mode=2
+[03:29:09] [yelp] L0=100% (256 pages) mode=2
+[03:29:18] [yelp] === DONE ===
+[03:29:18] [amazon_m2] ws=341MB bufsz=15 K=8 R=9
+[03:29:18] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[03:30:43] [amazon_m2] L0=5% (192 pages) mode=2
+[03:30:50] [amazon_m2] L0=10% (384 pages) mode=2
+[03:30:57] [amazon_m2] L0=20% (768 pages) mode=2
+[03:31:03] [amazon_m2] L0=50% (1920 pages) mode=2
+[03:31:09] [amazon_m2] L0=100% (3840 pages) mode=2
+[03:31:15] [amazon_m2] === DONE ===
+[03:31:15] [gowalla] ws=313MB bufsz=14 K=5 R=1
+[03:31:15] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[03:32:31] [gowalla] L0=5% (179 pages) mode=2
+[03:32:42] [gowalla] L0=10% (358 pages) mode=2
+[03:32:52] [gowalla] L0=20% (716 pages) mode=2
+[03:33:02] [gowalla] L0=50% (1792 pages) mode=2
+[03:33:11] [gowalla] L0=100% (3584 pages) mode=2
+[03:33:22] [gowalla] === DONE ===
+[03:33:22] === ALL DONE ===

mode15_ablation/run_20260501_023845.log

@@ -0,0 +1,5 @@
+[02:38:45] === START Mode 15 ablation ===
+[02:38:45] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451 K=14 R=1
+[02:38:45] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[02:40:12] [criteo_kaggle] mode=2

mode15_watchdog.log

@@ -0,0 +1,15 @@
+[02:17:41] mode15 watchdog start, waiting for taobao_n17...
+[02:38:41] taobao_n17 done, scp updated binary...
+      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+race_trace_v3.c: In function ‘remap_load’:
+race_trace_v3.c:102:34: warning: ignoring return value of ‘fread’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
+  102 |     char *jbuf = malloc(sz + 1); fread(jbuf, 1, sz, f); jbuf[sz] = 0; fclose(f);
+      |                                  ^~~~~~~~~~~~~~~~~~~~~
+Modes: 0=ALL-CXL.MEM 1=E 2=E+L0 3=ALL-NAND 4=HOST-CACHE 5=DPR-MODEL-UB 6=EWMA 7=HOST-L0+NAND 8=ALL-CXL.MEM-CAL 9=DPR-EWMA-POLLUTION 10=ALL-HOST-DRAM 11=DPR-EWMA-Q8 12=DPR-EWMA-Q4 13=DPR-EWMA-L0FIX 15=DPR-KRE-FUSION (K||R||EWMA + F1)
+[02:38:45] launching mode15 ablation...
+[02:38:45] === START Mode 15 ablation ===
+[02:38:45] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451 K=14 R=1
+[02:38:45] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[02:40:12] [criteo_kaggle] mode=2
+[02:40:36] mode15 watchdog done.

mode16_ablation/run_20260501_033353.log

@@ -0,0 +1,74 @@
+[03:33:53] === START Mode 16 (LFU-rank) ablation ===
+[03:33:53] [criteo_kaggle] ws=5380 bufsz=252 l0=6451 K=21 R=1 hot=0.01 cold=0.80
+[03:33:53] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[03:35:19] [criteo_kaggle] mode=2
+[03:36:17] [criteo_kaggle] mode=9
+[03:36:52] [criteo_kaggle] mode=13
+[03:37:27] [criteo_kaggle] mode=16
+[03:38:00] [criteo_kaggle] === DONE ===
+[03:38:00] [avazu] ws=2307 bufsz=108 l0=2764 K=298 R=1 hot=0.01 cold=0.80
+[03:38:00] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[03:39:26] [avazu] mode=2
+[03:39:34] [avazu] mode=9
+[03:39:40] [avazu] mode=13
+[03:39:46] [avazu] mode=16
+[03:39:52] [avazu] === DONE ===
+[03:39:52] [taobao] ws=102 bufsz=4 l0=102 K=38 R=1 hot=0.05 cold=0.50
+[03:39:52] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[03:41:17] [taobao] mode=2
+[03:41:25] [taobao] mode=9
+[03:41:31] [taobao] mode=13
+[03:41:37] [taobao] mode=16
+[03:41:43] [taobao] === DONE ===
+[03:41:43] [alibaba_ifashion] ws=532 bufsz=24 l0=614 K=10 R=1 hot=0.10 cold=0.30
+[03:41:43] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:43:09] [alibaba_ifashion] mode=2
+[03:43:50] [alibaba_ifashion] mode=9
+[03:45:34] [alibaba_ifashion] mode=13
+[03:47:23] [alibaba_ifashion] mode=16
+[03:51:09] [alibaba_ifashion] === DONE ===
+[03:51:09] [tmall] ws=531 bufsz=24 l0=614 K=14 R=1 hot=0.10 cold=0.30
+[03:51:09] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:52:35] [tmall] mode=2
+[03:52:45] [tmall] mode=9
+[03:52:53] [tmall] mode=13
+[03:53:01] [tmall] mode=16
+[03:53:09] [tmall] === DONE ===
+[03:53:09] [lastfm_1k] ws=309 bufsz=14 l0=358 K=31 R=1 hot=0.10 cold=0.30
+[03:53:09] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[03:54:25] [lastfm_1k] mode=2
+[03:54:39] [lastfm_1k] mode=9
+[03:56:57] [lastfm_1k] mode=13
+[03:59:16] [lastfm_1k] mode=16
+[04:01:57] [lastfm_1k] === DONE ===
+[04:01:57] [yelp] ws=37 bufsz=1 l0=25 K=5 R=1 hot=0.10 cold=0.30
+[04:01:57] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[04:03:19] [yelp] mode=2
+[04:03:27] [yelp] mode=9
+[04:03:33] [yelp] mode=13
+[04:03:39] [yelp] mode=16
+[04:03:46] [yelp] === DONE ===
+[04:03:46] [amazon_m2] ws=341 bufsz=15 l0=384 K=8 R=9 hot=0.15 cold=0.10
+[04:03:46] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[04:05:11] [amazon_m2] mode=2
+[04:05:18] [amazon_m2] mode=9
+[04:05:23] [amazon_m2] mode=13
+[04:05:28] [amazon_m2] mode=16
+[04:05:33] [amazon_m2] === DONE ===
+[04:05:33] [gowalla] ws=313 bufsz=14 l0=358 K=5 R=1 hot=0.15 cold=0.10
+[04:05:33] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[04:06:59] [gowalla] mode=2
+[04:07:09] [gowalla] mode=9
+[04:07:18] [gowalla] mode=13
+[04:07:27] [gowalla] mode=16
+[04:07:38] [gowalla] === DONE ===
+[04:07:38] === ALL DONE ===

mode16_watchdog.log

@@ -0,0 +1,84 @@
+[03:12:49] mode16 watchdog start, waiting for L0 sweep ALL_DONE...
+[03:33:49] post-mode15 chain done, scp updated binary...
+      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+race_trace_v3.c: In function ‘remap_load’:
+race_trace_v3.c:102:34: warning: ignoring return value of ‘fread’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
+  102 |     char *jbuf = malloc(sz + 1); fread(jbuf, 1, sz, f); jbuf[sz] = 0; fclose(f);
+      |                                  ^~~~~~~~~~~~~~~~~~~~~
+Modes: 0=ALL-CXL.MEM 1=E 2=E+L0 3=ALL-NAND 4=HOST-CACHE 5=DPR-MODEL-UB 6=EWMA 7=HOST-L0+NAND 8=ALL-CXL.MEM-CAL 9=DPR-EWMA-POLLUTION 10=ALL-HOST-DRAM 11=DPR-EWMA-Q8 12=DPR-EWMA-Q4 13=DPR-EWMA-L0FIX 15=DPR-KRE-FUSION 16=DPR-LFU-RANK (freq-rank static + R-rule middle)
+[03:33:53] launching Mode 16 ablation...
+[03:33:53] === START Mode 16 (LFU-rank) ablation ===
+[03:33:53] [criteo_kaggle] ws=5380 bufsz=252 l0=6451 K=21 R=1 hot=0.01 cold=0.80
+[03:33:53] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[03:35:19] [criteo_kaggle] mode=2
+[03:36:17] [criteo_kaggle] mode=9
+[03:36:52] [criteo_kaggle] mode=13
+[03:37:27] [criteo_kaggle] mode=16
+[03:38:00] [criteo_kaggle] === DONE ===
+[03:38:00] [avazu] ws=2307 bufsz=108 l0=2764 K=298 R=1 hot=0.01 cold=0.80
+[03:38:00] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[03:39:26] [avazu] mode=2
+[03:39:34] [avazu] mode=9
+[03:39:40] [avazu] mode=13
+[03:39:46] [avazu] mode=16
+[03:39:52] [avazu] === DONE ===
+[03:39:52] [taobao] ws=102 bufsz=4 l0=102 K=38 R=1 hot=0.05 cold=0.50
+[03:39:52] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[03:41:17] [taobao] mode=2
+[03:41:25] [taobao] mode=9
+[03:41:31] [taobao] mode=13
+[03:41:37] [taobao] mode=16
+[03:41:43] [taobao] === DONE ===
+[03:41:43] [alibaba_ifashion] ws=532 bufsz=24 l0=614 K=10 R=1 hot=0.10 cold=0.30
+[03:41:43] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:43:09] [alibaba_ifashion] mode=2
+[03:43:50] [alibaba_ifashion] mode=9
+[03:45:34] [alibaba_ifashion] mode=13
+[03:47:23] [alibaba_ifashion] mode=16
+[03:51:09] [alibaba_ifashion] === DONE ===
+[03:51:09] [tmall] ws=531 bufsz=24 l0=614 K=14 R=1 hot=0.10 cold=0.30
+[03:51:09] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:52:35] [tmall] mode=2
+[03:52:45] [tmall] mode=9
+[03:52:53] [tmall] mode=13
+[03:53:01] [tmall] mode=16
+[03:53:09] [tmall] === DONE ===
+[03:53:09] [lastfm_1k] ws=309 bufsz=14 l0=358 K=31 R=1 hot=0.10 cold=0.30
+[03:53:09] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[03:54:25] [lastfm_1k] mode=2
+[03:54:39] [lastfm_1k] mode=9
+[03:56:57] [lastfm_1k] mode=13
+[03:59:16] [lastfm_1k] mode=16
+[04:01:57] [lastfm_1k] === DONE ===
+[04:01:57] [yelp] ws=37 bufsz=1 l0=25 K=5 R=1 hot=0.10 cold=0.30
+[04:01:57] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[04:03:19] [yelp] mode=2
+[04:03:27] [yelp] mode=9
+[04:03:33] [yelp] mode=13
+[04:03:39] [yelp] mode=16
+[04:03:46] [yelp] === DONE ===
+[04:03:46] [amazon_m2] ws=341 bufsz=15 l0=384 K=8 R=9 hot=0.15 cold=0.10
+[04:03:46] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[04:05:11] [amazon_m2] mode=2
+[04:05:18] [amazon_m2] mode=9
+[04:05:23] [amazon_m2] mode=13
+[04:05:28] [amazon_m2] mode=16
+[04:05:33] [amazon_m2] === DONE ===
+[04:05:33] [gowalla] ws=313 bufsz=14 l0=358 K=5 R=1 hot=0.15 cold=0.10
+[04:05:33] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[04:06:59] [gowalla] mode=2
+[04:07:09] [gowalla] mode=9
+[04:07:18] [gowalla] mode=13
+[04:07:27] [gowalla] mode=16
+[04:07:38] [gowalla] === DONE ===
+[04:07:38] === ALL DONE ===
+[04:07:38] mode16 watchdog done.

mode17_ablation/run_20260501_213736.log

@@ -0,0 +1,83 @@
+[21:37:36] === START Mode 17 (DPR-LB) ablation ===
+[21:37:36] [criteo_kaggle] ws=5380 bufsz=252 l0=6451 K=21 R=1
+[21:37:36] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[21:39:06] [criteo_kaggle] mode=2
+[21:40:09] [criteo_kaggle] mode=9
+[21:40:44] [criteo_kaggle] mode=13
+[21:41:20] [criteo_kaggle] mode=16
+[21:41:54] [criteo_kaggle] mode=17
+[21:43:27] [criteo_kaggle] === DONE ===
+[21:43:27] [avazu] ws=2307 bufsz=108 l0=2764 K=298 R=1
+[21:43:27] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[21:44:54] [avazu] mode=2
+[21:45:02] [avazu] mode=9
+[21:45:08] [avazu] mode=13
+[21:45:14] [avazu] mode=16
+[21:45:20] [avazu] mode=17
+[21:45:27] [avazu] === DONE ===
+[21:45:27] [taobao] ws=102 bufsz=4 l0=102 K=38 R=1
+[21:45:27] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[21:46:53] [taobao] mode=2
+[21:46:58] [taobao] mode=9
+[21:47:02] [taobao] mode=13
+[21:47:07] [taobao] mode=16
+[21:47:12] [taobao] mode=17
+[21:47:16] [taobao] === DONE ===
+[21:47:16] [alibaba_ifashion] ws=532 bufsz=24 l0=614 K=10 R=1
+[21:47:16] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[21:48:42] [alibaba_ifashion] mode=2
+[21:48:48] [alibaba_ifashion] mode=9
+[21:48:53] [alibaba_ifashion] mode=13
+[21:48:58] [alibaba_ifashion] mode=16
+[21:49:03] [alibaba_ifashion] mode=17
+[21:49:08] [alibaba_ifashion] === DONE ===
+[21:49:08] [tmall] ws=531 bufsz=24 l0=614 K=14 R=1
+[21:49:08] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[21:50:34] [tmall] mode=2
+[21:50:44] [tmall] mode=9
+[21:50:52] [tmall] mode=13
+[21:51:01] [tmall] mode=16
+[21:51:09] [tmall] mode=17
+[21:51:27] [tmall] === DONE ===
+[21:51:27] [lastfm_1k] ws=309 bufsz=14 l0=358 K=31 R=1
+[21:51:27] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[21:52:53] [lastfm_1k] mode=2
+[21:53:07] [lastfm_1k] mode=9
+[21:55:26] [lastfm_1k] mode=13
+[21:57:48] [lastfm_1k] mode=16
+[22:00:33] [lastfm_1k] mode=17
+[22:03:14] [lastfm_1k] === DONE ===
+[22:03:14] [yelp] ws=37 bufsz=1 l0=25 K=5 R=1
+[22:03:14] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[22:04:40] [yelp] mode=2
+[22:04:49] [yelp] mode=9
+[22:04:55] [yelp] mode=13
+[22:05:02] [yelp] mode=16
+[22:05:09] [yelp] mode=17
+[22:05:20] [yelp] === DONE ===
+[22:05:20] [amazon_m2] ws=341 bufsz=15 l0=384 K=8 R=9
+[22:05:20] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[22:06:46] [amazon_m2] mode=2
+[22:06:53] [amazon_m2] mode=9
+[22:06:58] [amazon_m2] mode=13
+[22:07:03] [amazon_m2] mode=16
+[22:07:07] [amazon_m2] mode=17
+[22:07:12] [amazon_m2] === DONE ===
+[22:07:12] [gowalla] ws=313 bufsz=14 l0=358 K=5 R=1
+[22:07:12] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[22:08:38] [gowalla] mode=2
+[22:08:49] [gowalla] mode=9
+[22:08:59] [gowalla] mode=13
+[22:09:08] [gowalla] mode=16
+[22:09:19] [gowalla] mode=17
+[22:09:50] [gowalla] === DONE ===
+[22:09:50] === ALL DONE ===

mode19_ablation/run_20260502_003823.log

@@ -0,0 +1,65 @@
+[00:38:23] === START Mode 19 R-only ablation ===
+[00:38:23] [criteo_kaggle] ws=5380 bufsz=252 l0=6451 R_P50=2 R_P75=6
+[00:38:23] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[00:39:49] [criteo_kaggle] R_thr=1 (R1)
+[00:40:25] [criteo_kaggle] R_thr=2 (R50)
+[00:40:58] [criteo_kaggle] R_thr=6 (R75)
+[00:41:33] [criteo_kaggle] === DONE ===
+[00:41:33] [avazu] ws=2307 bufsz=108 l0=2764 R_P50=2 R_P75=16
+[00:41:33] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[00:42:59] [avazu] R_thr=1 (R1)
+[00:43:05] [avazu] R_thr=2 (R50)
+[00:43:11] [avazu] R_thr=16 (R75)
+[00:43:17] [avazu] === DONE ===
+[00:43:17] [taobao] ws=102 bufsz=4 l0=102 R_P50=2 R_P75=8
+[00:43:17] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[00:44:38] [taobao] R_thr=1 (R1)
+[00:44:44] [taobao] R_thr=2 (R50)
+[00:44:51] [taobao] R_thr=8 (R75)
+[00:44:59] [taobao] === DONE ===
+[00:44:59] [alibaba_ifashion] ws=532 bufsz=24 l0=614 R_P50=3 R_P75=7
+[00:44:59] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[00:46:25] [alibaba_ifashion] R_thr=1 (R1)
+[00:47:53] [alibaba_ifashion] R_thr=3 (R50)
+[00:50:45] [alibaba_ifashion] R_thr=7 (R75)
+[00:56:13] [alibaba_ifashion] === DONE ===
+[00:56:13] [tmall] ws=531 bufsz=24 l0=614 R_P50=147 R_P75=449
+[00:56:13] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[00:57:39] [tmall] R_thr=1 (R1)
+[00:57:45] [tmall] R_thr=147 (R50)
+[00:58:00] [tmall] R_thr=449 (R75)
+[00:58:27] [tmall] === DONE ===
+[00:58:27] [lastfm_1k] ws=309 bufsz=14 l0=358 R_P50=2 R_P75=4
+[00:58:27] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[00:59:48] [lastfm_1k] R_thr=1 (R1)
+[01:00:10] [lastfm_1k] R_thr=2 (R50)
+[01:00:41] [lastfm_1k] R_thr=4 (R75)
+[01:01:30] [lastfm_1k] === DONE ===
+[01:01:30] [yelp] ws=37 bufsz=1 l0=25 R_P50=87 R_P75=228
+[01:01:30] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[01:02:57] [yelp] R_thr=1 (R1)
+[01:03:02] [yelp] R_thr=87 (R50)
+[01:03:21] [yelp] R_thr=228 (R75)
+[01:03:50] [yelp] === DONE ===
+[01:03:50] [amazon_m2] ws=341 bufsz=15 l0=384 R_P50=995 R_P75=2320
+[01:03:50] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[01:05:11] [amazon_m2] R_thr=1 (R1)
+[01:05:16] [amazon_m2] R_thr=995 (R50)
+[01:05:21] [amazon_m2] R_thr=2320 (R75)
+[01:05:26] [amazon_m2] === DONE ===
+[01:05:26] [gowalla] ws=313 bufsz=14 l0=358 R_P50=111 R_P75=367
+[01:05:26] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[01:06:52] [gowalla] R_thr=1 (R1)
+[01:06:58] [gowalla] R_thr=111 (R50)
+[01:07:39] [gowalla] R_thr=367 (R75)
+[01:09:01] [gowalla] === DONE ===
+[01:09:01] === ALL DONE ===

mode19_watchdog.log

@@ -0,0 +1,70 @@
+[00:30:18] mode19 watchdog start, waiting for mode5 fairness ALL_DONE...
+[00:38:18] mode5 done, scp updated binary...
+Modes: 0=ALL-CXL.MEM 1=E 2=E+L0 3=ALL-NAND 4=HOST-CACHE 5=DPR-MODEL-UB 6=EWMA 7=HOST-L0+NAND 8=ALL-CXL.MEM-CAL 9=DPR-EWMA-POLLUTION 10=ALL-HOST-DRAM 11=DPR-EWMA-Q8 12=DPR-EWMA-Q4 13=DPR-EWMA-L0FIX 15=DPR-KRE-FUSION 16=DPR-LFU-RANK 17=DPR-LB 18=DPR-LB-OOO 19=DPR-R-ONLY (modeled NAND, R-rule only)
+[00:38:23] launching Mode 19 ablation...
+[00:38:23] === START Mode 19 R-only ablation ===
+[00:38:23] [criteo_kaggle] ws=5380 bufsz=252 l0=6451 R_P50=2 R_P75=6
+[00:38:23] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[00:39:49] [criteo_kaggle] R_thr=1 (R1)
+[00:40:25] [criteo_kaggle] R_thr=2 (R50)
+[00:40:58] [criteo_kaggle] R_thr=6 (R75)
+[00:41:33] [criteo_kaggle] === DONE ===
+[00:41:33] [avazu] ws=2307 bufsz=108 l0=2764 R_P50=2 R_P75=16
+[00:41:33] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[00:42:59] [avazu] R_thr=1 (R1)
+[00:43:05] [avazu] R_thr=2 (R50)
+[00:43:11] [avazu] R_thr=16 (R75)
+[00:43:17] [avazu] === DONE ===
+[00:43:17] [taobao] ws=102 bufsz=4 l0=102 R_P50=2 R_P75=8
+[00:43:17] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[00:44:38] [taobao] R_thr=1 (R1)
+[00:44:44] [taobao] R_thr=2 (R50)
+[00:44:51] [taobao] R_thr=8 (R75)
+[00:44:59] [taobao] === DONE ===
+[00:44:59] [alibaba_ifashion] ws=532 bufsz=24 l0=614 R_P50=3 R_P75=7
+[00:44:59] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[00:46:25] [alibaba_ifashion] R_thr=1 (R1)
+[00:47:53] [alibaba_ifashion] R_thr=3 (R50)
+[00:50:45] [alibaba_ifashion] R_thr=7 (R75)
+[00:56:13] [alibaba_ifashion] === DONE ===
+[00:56:13] [tmall] ws=531 bufsz=24 l0=614 R_P50=147 R_P75=449
+[00:56:13] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[00:57:39] [tmall] R_thr=1 (R1)
+[00:57:45] [tmall] R_thr=147 (R50)
+[00:58:00] [tmall] R_thr=449 (R75)
+[00:58:27] [tmall] === DONE ===
+[00:58:27] [lastfm_1k] ws=309 bufsz=14 l0=358 R_P50=2 R_P75=4
+[00:58:27] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[00:59:48] [lastfm_1k] R_thr=1 (R1)
+[01:00:10] [lastfm_1k] R_thr=2 (R50)
+[01:00:41] [lastfm_1k] R_thr=4 (R75)
+[01:01:30] [lastfm_1k] === DONE ===
+[01:01:30] [yelp] ws=37 bufsz=1 l0=25 R_P50=87 R_P75=228
+[01:01:30] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[01:02:57] [yelp] R_thr=1 (R1)
+[01:03:02] [yelp] R_thr=87 (R50)
+[01:03:21] [yelp] R_thr=228 (R75)
+[01:03:50] [yelp] === DONE ===
+[01:03:50] [amazon_m2] ws=341 bufsz=15 l0=384 R_P50=995 R_P75=2320
+[01:03:50] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[01:05:11] [amazon_m2] R_thr=1 (R1)
+[01:05:16] [amazon_m2] R_thr=995 (R50)
+[01:05:21] [amazon_m2] R_thr=2320 (R75)
+[01:05:26] [amazon_m2] === DONE ===
+[01:05:26] [gowalla] ws=313 bufsz=14 l0=358 R_P50=111 R_P75=367
+[01:05:26] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[01:06:52] [gowalla] R_thr=1 (R1)
+[01:06:58] [gowalla] R_thr=111 (R50)
+[01:07:39] [gowalla] R_thr=367 (R75)
+[01:09:01] [gowalla] === DONE ===
+[01:09:01] === ALL DONE ===
+[01:09:01] mode19 watchdog done.

mode_matrix_opt1_kr/run_20260428_000822.log

@@ -0,0 +1,86 @@
+[00:08:22] === START (Opt 1: K-R budget split β=0.7) ===
+[00:08:22] [amazon_Electronics] ws=63MB bufsz=2 l0=51 K=13 R=1
+[00:08:22] [amazon_Electronics] reboot VM ssd=6144 bufsz=2MB
+/dev/dax0.0
+[00:09:51] [amazon_Electronics] mode=2 K=13 R=1
+[00:09:59] [amazon_Electronics] mode=9 K=13 R=1
+[00:10:07] [amazon_Electronics] === DONE ===
+[00:10:07] [taobao] ws=102MB bufsz=4 l0=102 K=34 R=1
+[00:10:07] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[00:11:33] [taobao] mode=2 K=34 R=1
+[00:11:41] [taobao] mode=9 K=34 R=1
+[00:11:47] [taobao] === DONE ===
+[00:11:47] [foursquare] ws=25MB bufsz=1 l0=25 K=10 R=1
+[00:11:47] [foursquare] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[00:13:13] [foursquare] mode=2 K=10 R=1
+[00:13:23] [foursquare] mode=9 K=10 R=1
+[00:13:34] [foursquare] === DONE ===
+[00:13:34] [lastfm_1k] ws=309MB bufsz=14 l0=358 K=23 R=1
+[00:13:34] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[00:15:00] [lastfm_1k] mode=2 K=23 R=1
+[00:15:13] [lastfm_1k] mode=9 K=23 R=1
+[00:17:31] [lastfm_1k] === DONE ===
+[00:17:31] [bookcrossing] ws=81MB bufsz=3 l0=76 K=10 R=1
+[00:17:31] [bookcrossing] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[00:18:58] [bookcrossing] mode=2 K=10 R=1
+[00:19:05] [bookcrossing] mode=9 K=10 R=1
+[00:19:11] [bookcrossing] === DONE ===
+[00:19:11] [alibaba_ifashion] ws=532MB bufsz=24 l0=614 K=9 R=1
+[00:19:11] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[00:20:37] [alibaba_ifashion] mode=2 K=9 R=1
+[00:21:16] [alibaba_ifashion] mode=9 K=9 R=1
+[00:22:58] [alibaba_ifashion] === DONE ===
+[00:22:58] [gowalla] ws=313MB bufsz=14 l0=358 K=5 R=1
+[00:22:58] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[00:24:24] [gowalla] mode=2 K=5 R=1
+[00:24:35] [gowalla] mode=9 K=5 R=1
+[00:24:44] [gowalla] === DONE ===
+[00:24:44] [retailrocket] ws=36MB bufsz=1 l0=25 K=12 R=1
+[00:24:44] [retailrocket] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[00:26:11] [retailrocket] mode=2 K=12 R=1
+[00:26:16] [retailrocket] mode=9 K=12 R=1
+[00:26:21] [retailrocket] === DONE ===
+[00:26:21] [yelp] ws=37MB bufsz=1 l0=25 K=5 R=1
+[00:26:21] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[00:27:49] [yelp] mode=2 K=5 R=1
+[00:27:58] [yelp] mode=9 K=5 R=1
+[00:28:04] [yelp] === DONE ===
+[00:28:04] [lastfm_360k] ws=67MB bufsz=3 l0=76 K=8 R=1
+[00:28:04] [lastfm_360k] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[00:29:31] [lastfm_360k] mode=2 K=8 R=1
+[00:29:40] [lastfm_360k] mode=9 K=8 R=1
+[00:29:45] [lastfm_360k] === DONE ===
+[00:29:45] [amazon_m2] ws=341MB bufsz=15 l0=384 K=8 R=6
+[00:29:45] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[00:31:07] [amazon_m2] mode=2 K=8 R=6
+[00:31:13] [amazon_m2] mode=9 K=8 R=6
+[00:31:18] [amazon_m2] === DONE ===
+[00:31:18] [tmall] ws=531MB bufsz=24 l0=614 K=13 R=1
+[00:31:18] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[00:32:44] [tmall] mode=2 K=13 R=1
+[00:32:50] [tmall] mode=9 K=13 R=1
+[00:32:55] [tmall] === DONE ===
+[00:32:55] [avazu] ws=2307MB bufsz=108 l0=2764 K=287 R=1
+[00:32:55] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[00:35:14] [avazu] mode=2 K=287 R=1
+[00:35:21] [avazu] mode=9 K=287 R=1
+[00:35:27] [avazu] === DONE ===
+[00:35:27] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451 K=15 R=1
+[00:35:27] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[00:36:54] [criteo_kaggle] mode=2 K=15 R=1
+[00:38:00] [criteo_kaggle] mode=9 K=15 R=1
+[00:38:35] [criteo_kaggle] === DONE ===
+[00:38:35] === ALL DONE ===

mode_matrix_per_ssd/run_20260426_173452.log

@@ -0,0 +1,86 @@
+[17:34:52] === START ===
+[17:34:52] [lastfm_1k] ws=309MB ssd_fixed=6144MB | dev_dram=14MB | l0=358p (~1MB)
+[17:34:52] [lastfm_1k] rebooting VM with ssd_size=6144MB bufsz=14MB
+/dev/dax0.0
+[17:36:18] [lastfm_1k] probing race_trace_v3...
+[17:36:20] [lastfm_1k] probe OK
+[17:36:20] [lastfm_1k] mode=0 cap=992 l0=358
+[17:41:17] [lastfm_1k] mode=2 cap=992 l0=358
+[17:41:56] [lastfm_1k] mode=3 cap=992 l0=358
+[17:42:10] [lastfm_1k] mode=4 cap=992 l0=358
+[17:47:09] [lastfm_1k] mode=9 cap=992 l0=358
+[17:49:28] [lastfm_1k] mode=10 cap=992 l0=358
+[17:50:18] [lastfm_1k] === DONE ===
+[17:50:18] [gowalla] ws=313MB ssd_fixed=6144MB | dev_dram=14MB | l0=358p (~1MB)
+[17:50:18] [gowalla] rebooting VM with ssd_size=6144MB bufsz=14MB
+/dev/dax0.0
+[17:51:45] [gowalla] probing race_trace_v3...
+[17:51:45] [gowalla] probe OK
+[17:51:46] [gowalla] mode=0 cap=2000 l0=358
+[17:54:09] [gowalla] mode=2 cap=2000 l0=358
+[17:54:20] [gowalla] mode=3 cap=2000 l0=358
+[17:54:31] [gowalla] mode=4 cap=2000 l0=358
+[17:56:54] [gowalla] mode=9 cap=2000 l0=358
+[17:57:04] [gowalla] mode=10 cap=2000 l0=358
+[17:57:48] [gowalla] === DONE ===
+[17:57:48] [amazon_m2] ws=341MB ssd_fixed=6144MB | dev_dram=15MB | l0=384p (~1MB)
+[17:57:48] [amazon_m2] rebooting VM with ssd_size=6144MB bufsz=15MB
+/dev/dax0.0
+[17:59:14] [amazon_m2] probing race_trace_v3...
+[17:59:15] [amazon_m2] probe OK
+[17:59:15] [amazon_m2] mode=0 cap=2000 l0=384
+[17:59:23] [amazon_m2] mode=2 cap=2000 l0=384
+[17:59:29] [amazon_m2] mode=3 cap=2000 l0=384
+[17:59:35] [amazon_m2] mode=4 cap=2000 l0=384
+[17:59:43] [amazon_m2] mode=9 cap=2000 l0=384
+[17:59:47] [amazon_m2] mode=10 cap=2000 l0=384
+[18:06:53] [amazon_m2] === DONE ===
+[18:06:53] [tmall] ws=531MB ssd_fixed=6144MB | dev_dram=24MB | l0=614p (~2MB)
+[18:06:53] [tmall] rebooting VM with ssd_size=6144MB bufsz=24MB
+/dev/dax0.0
+[18:09:16] [tmall] probing race_trace_v3...
+[18:09:17] [tmall] probe OK
+[18:09:18] [tmall] mode=0 cap=2000 l0=614
+[18:10:30] [tmall] mode=2 cap=2000 l0=614
+[18:10:39] [tmall] mode=3 cap=2000 l0=614
+[18:10:49] [tmall] mode=4 cap=2000 l0=614
+[18:12:00] [tmall] mode=9 cap=2000 l0=614
+[18:12:08] [tmall] mode=10 cap=2000 l0=614
+[18:13:21] [tmall] === DONE ===
+[18:13:21] [alibaba_ifashion] ws=532MB ssd_fixed=6144MB | dev_dram=24MB | l0=614p (~2MB)
+[18:13:21] [alibaba_ifashion] rebooting VM with ssd_size=6144MB bufsz=24MB
+/dev/dax0.0
+[18:14:42] [alibaba_ifashion] probing race_trace_v3...
+[18:14:43] [alibaba_ifashion] probe OK
+[18:14:43] [alibaba_ifashion] mode=0 cap=7805 l0=614
+[18:14:48] [alibaba_ifashion] mode=2 cap=7805 l0=614
+[18:14:53] [alibaba_ifashion] mode=3 cap=7805 l0=614
+[18:14:57] [alibaba_ifashion] mode=4 cap=7805 l0=614
+[18:15:02] [alibaba_ifashion] mode=9 cap=7805 l0=614
+[18:15:07] [alibaba_ifashion] mode=10 cap=7805 l0=614
+[18:15:12] [alibaba_ifashion] === DONE ===
+[18:15:12] [avazu] ws=2307MB ssd_fixed=6144MB | dev_dram=108MB | l0=2764p (~10MB)
+[18:15:12] [avazu] rebooting VM with ssd_size=6144MB bufsz=108MB
+/dev/dax0.0
+[18:16:37] [avazu] probing race_trace_v3...
+[18:16:38] [avazu] probe OK
+[18:16:39] [avazu] mode=0 cap=2000 l0=2764
+[18:16:45] [avazu] mode=2 cap=2000 l0=2764
+[18:16:52] [avazu] mode=3 cap=2000 l0=2764
+[18:17:00] [avazu] mode=4 cap=2000 l0=2764
+[18:17:05] [avazu] mode=9 cap=2000 l0=2764
+[18:17:11] [avazu] mode=10 cap=2000 l0=2764
+[18:34:08] [avazu] === DONE ===
+[18:34:08] [criteo_kaggle] ws=5380MB ssd_fixed=6144MB | dev_dram=252MB | l0=6451p (~25MB)
+[18:34:08] [criteo_kaggle] rebooting VM with ssd_size=6144MB bufsz=252MB
+/dev/dax0.0
+[18:39:32] [criteo_kaggle] probing race_trace_v3...
+[18:39:35] [criteo_kaggle] probe OK
+[18:39:35] [criteo_kaggle] mode=0 cap=10000 l0=6451
+[18:43:30] [criteo_kaggle] mode=2 cap=10000 l0=6451
+[18:44:41] [criteo_kaggle] mode=3 cap=10000 l0=6451
+[18:45:44] [criteo_kaggle] mode=4 cap=10000 l0=6451
+[18:49:01] [criteo_kaggle] mode=9 cap=10000 l0=6451
+[18:49:35] [criteo_kaggle] mode=10 cap=10000 l0=6451
+[19:06:32] [criteo_kaggle] === DONE ===
+[19:06:32] === ALL DONE ===

mode_matrix_per_ssd/run_20260426_190751.log

@@ -0,0 +1,14 @@
+[19:07:51] === START ===
+[19:07:51] [alibaba_ifashion] ws=532MB ssd_fixed=6144MB | dev_dram=24MB | l0=614p (~2MB)
+[19:07:51] [alibaba_ifashion] rebooting VM with ssd_size=6144MB bufsz=24MB
+/dev/dax0.0
+[19:13:05] [alibaba_ifashion] probing race_trace_v3...
+[19:13:06] [alibaba_ifashion] probe OK
+[19:13:07] [alibaba_ifashion] mode=0 cap=7805 l0=614
+[19:38:09] [alibaba_ifashion] mode=2 cap=7805 l0=614
+[20:03:10] [alibaba_ifashion] mode=3 cap=7805 l0=614
+[20:05:19] [alibaba_ifashion] mode=4 cap=7805 l0=614
+[20:30:20] [alibaba_ifashion] mode=9 cap=7805 l0=614
+[20:46:02] [alibaba_ifashion] mode=10 cap=7805 l0=614
+[21:00:59] [alibaba_ifashion] === DONE ===
+[21:00:59] === ALL DONE ===

mode_matrix_per_ssd/summary.csv

@@ -0,0 +1,8 @@
+dataset,mode0_mean_ms,mode0_cv,mode0_hit_rate,mode2_mean_ms,mode2_cv,mode2_hit_rate,mode3_mean_ms,mode3_cv,mode3_hit_rate,mode4_mean_ms,mode4_cv,mode4_hit_rate,mode9_mean_ms,mode9_cv,mode9_hit_rate,mode10_mean_ms,mode10_cv,mode10_hit_rate,bufsz_mb,l0_pages,total_pg
+criteo_kaggle,4.018,3.9,0.0,0.737,2.1,91.1,0.589,2.3,0.0,3.270,0.7,71.2,0.588,2.1,100.0,NA,NA,NA,252,6451,10000
+avazu,0.083,220.2,0.0,0.227,1.5,100.0,0.445,0.5,0.0,0.005,1.8,99.8,0.046,0.2,100.0,NA,NA,NA,108,2764,1000
+amazon_m2,0.290,11.3,0.0,0.160,5.1,45.0,0.166,6.8,0.0,0.273,0.1,24.6,0.044,11.2,50.4,NA,NA,NA,15,384,2000
+alibaba_ifashion,NA,NA,NA,17.551,44.2,77.3,1.073,1.5,0.0,NA,NA,NA,NA,NA,NA,NA,NA,NA,24,614,2000
+tmall,6.559,1.6,0.0,0.426,1.5,82.9,0.416,2.3,0.0,6.489,0.1,22.3,0.734,4.5,38.2,0.004,7.2,0.0,24,614,2000
+gowalla,13.735,0.5,0.0,0.495,5.6,73.7,0.436,5.9,0.0,13.711,0.3,12.2,1.223,1.2,36.4,0.006,5.7,0.0,14,358,2000
+lastfm_1k,243.178,0.6,0.0,22.807,1.0,70.0,1.297,2.9,0.0,244.449,1.3,11.2,116.030,1.0,11.4,0.092,13.2,0.0,14,358,233

mode_matrix_per_ssd/summary.md

@@ -0,0 +1,82 @@
+# Per-SSD ratio Mode Matrix — 7 datasets × 6 modes
+
+**Configuration**(對齊 paper 主張之 cache budget):
+- SSD = 6144 MB(固定,FEMU geometry stable)
+- CXL-SSD device DRAM(`bufsz_mb`)= working_set × 48/1024
+- Host L0 = device DRAM × 10% = working_set × 0.0046875
+
+每 dataset 重啟 VM 應用對應 bufsz。Mode 矩陣每 mode 5 runs。
+
+## 主結果
+
+| Dataset | bufsz_MB | L0_MB | Mode 0(CXL.MEM)| **Mode 2(E+L0)** | Mode 3(NAND)| Mode 4(LRU)| Mode 9(EWMA θ\*)| Mode 10(host DRAM)|
+|---|---:|---:|---:|---:|---:|---:|---:|---:|
+| **criteo_kaggle** ⭐ | 252 | 25.2 | 4.02 ms | **0.74 ms**(91% hit)| 0.59 | 3.27 | 0.59(100% hit)| NA |
+| **avazu** | 108 | 10.8 | 0.08 | 0.23 | 0.45 | 0.005(99.8% hit)| 0.05(100%)| NA |
+| **amazon_m2** | 15 | 1.6 | 0.29 | 0.16 | 0.17 | 0.27 | 0.04(50%)| NA |
+| alibaba_ifashion(部分)| 24 | 2.5 | TIMEOUT | 17.55(77%)| **1.07** | TIMEOUT | TIMEOUT | NA |
+| **tmall** | 24 | 2.5 | 6.56 | **0.43**(83%)| 0.42 | 6.49 | 0.73 | < 0.05 |
+| **gowalla** | 14 | 1.4 | 13.74 | **0.50**(74%)| 0.44 | 13.71 | 1.22 | < 0.05 |
+| **lastfm_1k** | 14 | 1.4 | 243.18 | 22.81 | **1.30** | 244.45 | 116.03 | 0.09 |
+
+## 加速比(Mode 0 vs Mode 2 / Mode 9)
+
+| Dataset | E+L0 加速 | DPR-EWMA 加速 | 主要解讀 |
+|---|---:|---:|---|
+| criteo_kaggle | **5.5×** | 6.8× | Multi-table + Pareto top 5%→92% volume,K-routing 主結果 |
+| avazu | n/a | n/a | working set 0.6 GB / cache 11 MB → fit-in-cache,差異小 |
+| amazon_m2 | n/a | n/a | K_P95=3,K-routing 訊號弱,差異不顯著 |
+| alibaba_ifashion | n/a | n/a | **24 MB cache 對 532 MB ws thrash;K-routing 反而 17.55 ms 不如 NAND 1.07 ms** |
+| tmall | **15.4×** | 8.9× | Multi-routing 顯著 |
+| gowalla | **27.5×** | 11.2× | Multi-routing 極顯著 |
+| lastfm_1k | **10.7×** | 2.1× | Cache pressure 極大;EWMA 在此 dataset 收益較小 |
+
+## 重要觀察
+
+### 1. K-routing 之 multi-table workload 主結果(criteo / tmall / gowalla)
+
+per-SSD ratio 配置下,加速比較 fixed L0=16MB 之 mode_matrix 顯著:
+- criteo: 2.83× → **5.5×**
+- gowalla: 顯示 27.5× 為最強加速
+
+### 2. K-routing 之適用邊界(alibaba)
+
+24 MB cache 對 532 MB working set 之 thrash 場景下,K-routing 規則本身會引入 cache pollution overhead → Mode 2(17.55 ms)反而**比 NAND parallel(1.07 ms)更慢**。
+
+paper 應寫:「K-routing 規則之有效性需 cache 容量 ≥ working_set × 4.7%(48:1024 paper 比例),對 working set 過大或 K_P95 < 5 之 dataset 退化為純 LFU 或 NAND-parallel」。
+
+### 3. Mode 10(absolute host DRAM upper bound)— sub-50 μs/batch
+
+對小 dataset(tmall / gowalla / lastfm_1k)之 Mode 10 = < 0.05 ms,證明:**resident host DRAM 為 absolute ceiling,paper 之 E+L0 距此 ceiling 約 10× (criteo)**。
+
+### 4. EWMA cost-eq θ\*(Mode 9)接近 Mode 2(E+L0)best
+
+對 criteo,Mode 9 = 0.59 ms 與 Mode 2 之 0.74 ms 接近(略快因 100% hit rate)。Mode 9 之 cost-equation 自適應 routing 在 paper 主結果是另一條 valid path。
+
+## 對應 K/R 推導之 α / γ 自洽性
+
+對 criteo(per-SSD config):
+- N_acc = 113,045
+- N_L0 = 6,604
+- α = N_L0 / N_acc = **5.84%** ≈ 0.05 慣例(Path B confirmed)
+- L_cache(從 Mode 2 hit rate 91% 反算)≈ 4 batches,γ = 1−e^(−6/4) = 77.7% ≈ 0.75 慣例(Path C confirmed)
+
+詳推導見 `docs/method/kr_derivation.md`。
+
+## 已知 caveats
+
+1. **Mode 10 對 criteo / avazu / amazon_m2 = NA**:script 在某些 dataset 之 Mode 10 沒跑完(可能 memcpy 6 GB 時 SSH timeout,或 script 中斷)。實測對小 dataset Mode 10 ≈ 0.05 ms(host DRAM ~80ns/page × ~700 pages = 56 μs / batch)。
+2. **alibaba_ifashion mode 0/4/9/10 = TIMEOUT**:24 MB cache 對 532 MB working set 之 thrash 使 batch latency > 150 ms,5 min × 5 runs > 25 min 超 SSH timeout。**這正是 paper 之 cache-pressure 適用邊界證據**。
+3. **lastfm_1k Mode 0 = 243 ms** 為合理之 cache-thrash baseline(working set / cache = 22×)。
+
+## 檔案位置
+
+```
+research_data/results/mode_matrix_per_ssd/
+├── summary.csv          # 7 dataset × 6 mode matrix
+├── summary.md           # 本檔
+├── <dataset>/mode<m>/
+│   ├── <dataset>_mode<m>_<TS>.csv     # 5 runs
+│   └── <dataset>_mode<m>_<TS>_run<i>.log
+└── DONE.marker          # per-dataset
+```

mode_matrix_per_ssd/summary_per_ssd.csv

@@ -0,0 +1,8 @@
+dataset,mode0_mean_ms,mode0_cv,mode0_hit_rate,mode2_mean_ms,mode2_cv,mode2_hit_rate,mode3_mean_ms,mode3_cv,mode3_hit_rate,mode4_mean_ms,mode4_cv,mode4_hit_rate,mode9_mean_ms,mode9_cv,mode9_hit_rate,mode10_mean_ms,mode10_cv,mode10_hit_rate,bufsz_mb,l0_pages,total_pg
+criteo_kaggle,4.018,3.9,0.0,0.737,2.1,91.1,0.589,2.3,0.0,3.270,0.7,71.2,0.588,2.1,100.0,NA,NA,NA,252,6451,10000
+avazu,0.083,220.2,0.0,0.227,1.5,100.0,0.445,0.5,0.0,0.005,1.8,99.8,0.046,0.2,100.0,NA,NA,NA,108,2764,1000
+amazon_m2,0.290,11.3,0.0,0.160,5.1,45.0,0.166,6.8,0.0,0.273,0.1,24.6,0.044,11.2,50.4,NA,NA,NA,15,384,2000
+alibaba_ifashion,NA,NA,NA,17.551,44.2,77.3,1.073,1.5,0.0,NA,NA,NA,NA,NA,NA,NA,NA,NA,24,614,2000
+tmall,6.559,1.6,0.0,0.426,1.5,82.9,0.416,2.3,0.0,6.489,0.1,22.3,0.734,4.5,38.2,0.004,7.2,0.0,24,614,2000
+gowalla,13.735,0.5,0.0,0.495,5.6,73.7,0.436,5.9,0.0,13.711,0.3,12.2,1.223,1.2,36.4,0.006,5.7,0.0,14,358,2000
+lastfm_1k,243.178,0.6,0.0,22.807,1.0,70.0,1.297,2.9,0.0,244.449,1.3,11.2,116.030,1.0,11.4,0.092,13.2,0.0,14,358,233

mode_matrix_perds_kr/run_20260427_004033.log

@@ -0,0 +1,3 @@
+[00:40:33] === START (per-dataset K/R) ===
+[00:40:33] [amazon_Electronics] ws=63MB bufsz=2 l0=51 K=3 R=6
+[00:40:33] [amazon_Electronics] reboot VM ssd=6144 bufsz=2MB

post_mode15_chain.log

@@ -0,0 +1,188 @@
+[02:41:56] === post-mode15 chain start ===
+[02:41:56] Step 1: Cross-trace K/R transfer...
+[02:41:56] === START Cross-trace K/R transfer ===
+[02:41:56] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[02:41:56] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[02:43:17] [criteo_kaggle] src=src_criteo K=14 R=1
+[02:44:29] [criteo_kaggle] src=src_avazu K=283 R=1
+[02:45:29] [criteo_kaggle] src=src_taobao K=33 R=1
+[02:46:25] [criteo_kaggle] === DONE ===
+[02:46:25] [avazu] ws=2307MB bufsz=108 l0=2764
+[02:46:25] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[02:47:51] [avazu] src=src_criteo K=14 R=1
+[02:47:59] [avazu] src=src_avazu K=283 R=1
+[02:48:06] [avazu] src=src_taobao K=33 R=1
+[02:48:14] [avazu] === DONE ===
+[02:48:14] [amazon_m2] ws=341MB bufsz=15 l0=384
+[02:48:14] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[02:49:30] [amazon_m2] src=src_criteo K=14 R=1
+[02:49:37] [amazon_m2] src=src_avazu K=283 R=1
+[02:49:43] [amazon_m2] src=src_taobao K=33 R=1
+[02:49:50] [amazon_m2] === DONE ===
+[02:49:50] [alibaba_ifashion] ws=532MB bufsz=24 l0=614
+[02:49:50] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[02:51:11] [alibaba_ifashion] src=src_criteo K=14 R=1
+[02:51:49] [alibaba_ifashion] src=src_avazu K=283 R=1
+[02:52:26] [alibaba_ifashion] src=src_taobao K=33 R=1
+[02:53:05] [alibaba_ifashion] === DONE ===
+[02:53:05] [tmall] ws=531MB bufsz=24 l0=614
+[02:53:05] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[02:54:21] [tmall] src=src_criteo K=14 R=1
+[02:54:31] [tmall] src=src_avazu K=283 R=1
+[02:54:41] [tmall] src=src_taobao K=33 R=1
+[02:54:51] [tmall] === DONE ===
+[02:54:51] [gowalla] ws=313MB bufsz=14 l0=358
+[02:54:51] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[02:56:07] [gowalla] src=src_criteo K=14 R=1
+[02:56:17] [gowalla] src=src_avazu K=283 R=1
+[02:56:27] [gowalla] src=src_taobao K=33 R=1
+[02:56:37] [gowalla] === DONE ===
+[02:56:37] [lastfm_1k] ws=309MB bufsz=14 l0=358
+[02:56:37] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[02:58:03] [lastfm_1k] src=src_criteo K=14 R=1
+[02:58:08] [lastfm_1k] src=src_avazu K=283 R=1
+[02:58:13] [lastfm_1k] src=src_taobao K=33 R=1
+[02:58:17] [lastfm_1k] === DONE ===
+[02:58:17] [taobao] ws=102MB bufsz=4 l0=102
+[02:58:17] [taobao] reboot VM ssd=6144 bufsz=4MB
+/home/sadoo/Projects/kvcache/research_data/scripts/cylon_experiments/cross_trace_kr.sh: line 42: 2423781 Killed                  sudo pkill -9 -f qemu-system-x86_64
+/dev/dax0.0
+[03:00:33] [taobao] src=src_criteo K=14 R=1
+[03:00:41] [taobao] src=src_avazu K=283 R=1
+[03:00:49] [taobao] src=src_taobao K=33 R=1
+[03:00:56] [taobao] === DONE ===
+[03:00:56] [bookcrossing] ws=81MB bufsz=3 l0=76
+[03:00:56] [bookcrossing] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[03:02:17] [bookcrossing] src=src_criteo K=14 R=1
+[03:02:24] [bookcrossing] src=src_avazu K=283 R=1
+[03:02:31] [bookcrossing] src=src_taobao K=33 R=1
+[03:02:38] [bookcrossing] === DONE ===
+[03:02:38] [lastfm_360k] ws=67MB bufsz=3 l0=76
+[03:02:38] [lastfm_360k] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[03:03:59] [lastfm_360k] src=src_criteo K=14 R=1
+[03:04:08] [lastfm_360k] src=src_avazu K=283 R=1
+[03:04:17] [lastfm_360k] src=src_taobao K=33 R=1
+[03:04:26] [lastfm_360k] === DONE ===
+[03:04:26] [amazon_Electronics] ws=63MB bufsz=2 l0=51
+[03:04:26] [amazon_Electronics] reboot VM ssd=6144 bufsz=2MB
+/dev/dax0.0
+[03:05:51] [amazon_Electronics] src=src_criteo K=14 R=1
+[03:05:58] [amazon_Electronics] src=src_avazu K=283 R=1
+[03:06:06] [amazon_Electronics] src=src_taobao K=33 R=1
+[03:06:13] [amazon_Electronics] === DONE ===
+[03:06:13] [yelp] ws=37MB bufsz=1 l0=25
+[03:06:13] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[03:07:39] [yelp] src=src_criteo K=14 R=1
+[03:07:47] [yelp] src=src_avazu K=283 R=1
+[03:07:55] [yelp] src=src_taobao K=33 R=1
+[03:08:04] [yelp] === DONE ===
+[03:08:04] [retailrocket] ws=36MB bufsz=1 l0=25
+[03:08:04] [retailrocket] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[03:09:25] [retailrocket] src=src_criteo K=14 R=1
+[03:09:32] [retailrocket] src=src_avazu K=283 R=1
+[03:09:37] [retailrocket] src=src_taobao K=33 R=1
+[03:09:43] [retailrocket] === DONE ===
+[03:09:43] [foursquare] ws=25MB bufsz=1 l0=25
+[03:09:43] [foursquare] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[03:11:10] [foursquare] src=src_criteo K=14 R=1
+[03:11:20] [foursquare] src=src_avazu K=283 R=1
+[03:11:30] [foursquare] src=src_taobao K=33 R=1
+[03:11:41] [foursquare] === DONE ===
+[03:11:41] === ALL DONE ===
+[03:11:41] Step 2: L0 size sweep...
+[03:11:41] === START L0 size sweep ===
+[03:11:41] [criteo_kaggle] ws=5380MB bufsz=252 K=21 R=1
+[03:11:41] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[03:13:07] [criteo_kaggle] L0=5% (3225 pages) mode=2
+[03:13:19] [criteo_kaggle] L0=10% (6451 pages) mode=2
+[03:13:30] [criteo_kaggle] L0=20% (12902 pages) mode=2
+[03:13:41] [criteo_kaggle] L0=50% (32256 pages) mode=2
+[03:13:52] [criteo_kaggle] L0=100% (64512 pages) mode=2
+[03:14:04] [criteo_kaggle] === DONE ===
+[03:14:04] [avazu] ws=2307MB bufsz=108 K=298 R=1
+[03:14:04] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[03:15:30] [avazu] L0=5% (1382 pages) mode=2
+[03:15:39] [avazu] L0=10% (2764 pages) mode=2
+[03:15:46] [avazu] L0=20% (5529 pages) mode=2
+[03:15:54] [avazu] L0=50% (13824 pages) mode=2
+[03:16:00] [avazu] L0=100% (27648 pages) mode=2
+[03:16:07] [avazu] === DONE ===
+[03:16:07] [taobao] ws=102MB bufsz=4 K=38 R=1
+[03:16:07] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[03:17:33] [taobao] L0=5% (51 pages) mode=2
+[03:17:40] [taobao] L0=10% (102 pages) mode=2
+[03:17:48] [taobao] L0=20% (204 pages) mode=2
+[03:17:55] [taobao] L0=50% (512 pages) mode=2
+[03:18:02] [taobao] L0=100% (1024 pages) mode=2
+[03:18:10] [taobao] === DONE ===
+[03:18:10] [alibaba_ifashion] ws=532MB bufsz=24 K=10 R=1
+[03:18:10] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:19:26] [alibaba_ifashion] L0=5% (307 pages) mode=2
+[03:20:03] [alibaba_ifashion] L0=10% (614 pages) mode=2
+[03:20:41] [alibaba_ifashion] L0=20% (1228 pages) mode=2
+[03:21:18] [alibaba_ifashion] L0=50% (3072 pages) mode=2
+[03:21:55] [alibaba_ifashion] L0=100% (6144 pages) mode=2
+[03:22:30] [alibaba_ifashion] === DONE ===
+[03:22:30] [tmall] ws=531MB bufsz=24 K=14 R=1
+[03:22:30] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[03:23:56] [tmall] L0=5% (307 pages) mode=2
+[03:24:07] [tmall] L0=10% (614 pages) mode=2
+[03:24:17] [tmall] L0=20% (1228 pages) mode=2
+[03:24:27] [tmall] L0=50% (3072 pages) mode=2
+[03:24:36] [tmall] L0=100% (6144 pages) mode=2
+[03:24:46] [tmall] === DONE ===
+[03:24:46] [lastfm_1k] ws=309MB bufsz=14 K=31 R=1
+[03:24:46] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[03:26:02] [lastfm_1k] L0=5% (179 pages) mode=2
+[03:26:17] [lastfm_1k] L0=10% (358 pages) mode=2
+[03:26:30] [lastfm_1k] L0=20% (716 pages) mode=2
+[03:26:45] [lastfm_1k] L0=50% (1792 pages) mode=2
+[03:26:59] [lastfm_1k] L0=100% (3584 pages) mode=2
+[03:27:14] [lastfm_1k] === DONE ===
+[03:27:14] [yelp] ws=37MB bufsz=1 K=5 R=1
+[03:27:14] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[03:28:35] [yelp] L0=5% (12 pages) mode=2
+[03:28:44] [yelp] L0=10% (25 pages) mode=2
+[03:28:52] [yelp] L0=20% (51 pages) mode=2
+[03:29:01] [yelp] L0=50% (128 pages) mode=2
+[03:29:09] [yelp] L0=100% (256 pages) mode=2
+[03:29:18] [yelp] === DONE ===
+[03:29:18] [amazon_m2] ws=341MB bufsz=15 K=8 R=9
+[03:29:18] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[03:30:43] [amazon_m2] L0=5% (192 pages) mode=2
+[03:30:50] [amazon_m2] L0=10% (384 pages) mode=2
+[03:30:57] [amazon_m2] L0=20% (768 pages) mode=2
+[03:31:03] [amazon_m2] L0=50% (1920 pages) mode=2
+[03:31:09] [amazon_m2] L0=100% (3840 pages) mode=2
+[03:31:15] [amazon_m2] === DONE ===
+[03:31:15] [gowalla] ws=313MB bufsz=14 K=5 R=1
+[03:31:15] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[03:32:31] [gowalla] L0=5% (179 pages) mode=2
+[03:32:42] [gowalla] L0=10% (358 pages) mode=2
+[03:32:52] [gowalla] L0=20% (716 pages) mode=2
+[03:33:02] [gowalla] L0=50% (1792 pages) mode=2
+[03:33:11] [gowalla] L0=100% (3584 pages) mode=2
+[03:33:22] [gowalla] === DONE ===
+[03:33:22] === ALL DONE ===
+[03:33:22] === post-mode15 chain ALL DONE ===

route_b_20260501_011918.log

@@ -0,0 +1,321 @@
+[01:19:18] === Route B orchestrator START ===
+[01:19:18] Step 1: waiting for ewma_quant_reverse ALL_DONE.marker...
+[01:20:48] reverse-order DONE
+[01:20:48] Step 2: scp updated race_trace_v3.c to VM...
+      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+race_trace_v3.c: In function ‘remap_load’:
+race_trace_v3.c:102:34: warning: ignoring return value of ‘fread’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
+  102 |     char *jbuf = malloc(sz + 1); fread(jbuf, 1, sz, f); jbuf[sz] = 0; fclose(f);
+      |                                  ^~~~~~~~~~~~~~~~~~~~~
+Modes: 0=ALL-CXL.MEM 1=E 2=E+L0 3=ALL-NAND 4=HOST-CACHE 5=DPR-MODEL-UB 6=EWMA 7=HOST-L0+NAND 8=ALL-CXL.MEM-CAL 9=DPR-EWMA-POLLUTION 10=ALL-HOST-DRAM 11=DPR-EWMA-Q8 12=DPR-EWMA-Q4 13=DPR-EWMA-L0FIX
+[01:20:52] Step 3: F1 ablation...
+[01:20:52] === START F1 Ablation (Mode 9 vs 13) ===
+[01:20:52] [taobao] ws=102MB bufsz=4 l0=102
+[01:20:52] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[01:22:18] [taobao] mode=9
+[01:22:24] [taobao] mode=13
+[01:22:30] [taobao] === DONE ===
+[01:22:30] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[01:22:30] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[01:23:51] [criteo_kaggle] mode=9
+[01:24:03] [criteo_kaggle] mode=13
+[01:24:14] [criteo_kaggle] === DONE ===
+[01:24:14] [avazu] ws=2307MB bufsz=108 l0=2764
+[01:24:14] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[01:25:40] [avazu] mode=9
+[01:25:46] [avazu] mode=13
+[01:25:52] [avazu] === DONE ===
+[01:25:52] === ALL DONE ===
+[01:25:52] Step 4: β fine sweep...
+[01:25:52] === START β fine sweep ===
+[01:25:52] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[01:25:52] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[01:27:09] [criteo_kaggle] β=0.50 → K=21 R=1
+[01:28:07] [criteo_kaggle] β=0.55 → K=19 R=1
+[01:29:05] [criteo_kaggle] β=0.60 → K=17 R=1
+[01:30:02] [criteo_kaggle] β=0.65 → K=16 R=1
+[01:31:04] [criteo_kaggle] β=0.70 → K=15 R=1
+[01:32:08] [criteo_kaggle] β=0.7311 → K=14 R=1
+[01:33:14] [criteo_kaggle] β=0.75 → K=14 R=1
+[01:34:21] [criteo_kaggle] β=0.80 → K=13 R=1
+[01:35:28] [criteo_kaggle] β=0.85 → K=12 R=1
+[01:36:37] [criteo_kaggle] β=0.90 → K=12 R=1
+[01:37:47] [criteo_kaggle] === DONE ===
+[01:37:47] [avazu] ws=2307MB bufsz=108 l0=2764
+[01:37:47] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[01:39:13] [avazu] β=0.50 → K=298 R=1
+[01:39:21] [avazu] β=0.55 → K=296 R=1
+[01:39:28] [avazu] β=0.60 → K=293 R=1
+[01:39:36] [avazu] β=0.65 → K=291 R=1
+[01:39:43] [avazu] β=0.70 → K=287 R=1
+[01:39:51] [avazu] β=0.7311 → K=283 R=1
+[01:39:58] [avazu] β=0.75 → K=275 R=1
+[01:40:06] [avazu] β=0.80 → K=261 R=1
+[01:40:14] [avazu] β=0.85 → K=260 R=1
+[01:40:21] [avazu] β=0.90 → K=259 R=1
+[01:40:28] [avazu] === DONE ===
+[01:40:28] [amazon_m2] ws=341MB bufsz=15 l0=384
+[01:40:28] [amazon_m2] reboot VM ssd=6144 bufsz=15MB
+/dev/dax0.0
+[01:41:54] [amazon_m2] β=0.50 → K=8 R=9
+[01:42:01] [amazon_m2] β=0.55 → K=8 R=8
+[01:42:08] [amazon_m2] β=0.60 → K=8 R=7
+[01:42:14] [amazon_m2] β=0.65 → K=8 R=6
+[01:42:21] [amazon_m2] β=0.70 → K=8 R=6
+[01:42:28] [amazon_m2] β=0.7311 → K=7 R=5
+[01:42:34] [amazon_m2] β=0.75 → K=7 R=5
+[01:42:41] [amazon_m2] β=0.80 → K=7 R=4
+[01:42:48] [amazon_m2] β=0.85 → K=7 R=3
+[01:42:54] [amazon_m2] β=0.90 → K=7 R=2
+[01:43:01] [amazon_m2] === DONE ===
+[01:43:01] [alibaba_ifashion] ws=532MB bufsz=24 l0=614
+[01:43:01] [alibaba_ifashion] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[01:44:27] [alibaba_ifashion] β=0.50 → K=10 R=1
+[01:45:05] [alibaba_ifashion] β=0.55 → K=9 R=1
+[01:45:43] [alibaba_ifashion] β=0.60 → K=9 R=1
+[01:46:22] [alibaba_ifashion] β=0.65 → K=9 R=1
+[01:47:02] [alibaba_ifashion] β=0.70 → K=9 R=1
+[01:47:41] [alibaba_ifashion] β=0.7311 → K=9 R=1
+[01:48:20] [alibaba_ifashion] β=0.75 → K=8 R=1
+[01:49:03] [alibaba_ifashion] β=0.80 → K=8 R=1
+[01:49:45] [alibaba_ifashion] β=0.85 → K=8 R=1
+[01:50:27] [alibaba_ifashion] β=0.90 → K=8 R=1
+[01:51:09] [alibaba_ifashion] === DONE ===
+[01:51:09] [tmall] ws=531MB bufsz=24 l0=614
+[01:51:09] [tmall] reboot VM ssd=6144 bufsz=24MB
+/dev/dax0.0
+[01:52:31] [tmall] β=0.50 → K=14 R=1
+[01:52:41] [tmall] β=0.55 → K=14 R=1
+[01:52:51] [tmall] β=0.60 → K=13 R=1
+[01:53:01] [tmall] β=0.65 → K=13 R=1
+[01:53:11] [tmall] β=0.70 → K=13 R=1
+[01:53:22] [tmall] β=0.7311 → K=12 R=1
+[01:53:32] [tmall] β=0.75 → K=12 R=1
+[01:53:42] [tmall] β=0.80 → K=12 R=1
+[01:53:51] [tmall] β=0.85 → K=12 R=1
+[01:54:02] [tmall] β=0.90 → K=11 R=1
+[01:54:12] [tmall] === DONE ===
+[01:54:12] [gowalla] ws=313MB bufsz=14 l0=358
+[01:54:12] [gowalla] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[01:55:38] [gowalla] β=0.50 → K=5 R=1
+[01:55:49] [gowalla] β=0.55 → K=5 R=1
+[01:55:59] [gowalla] β=0.60 → K=5 R=1
+[01:56:10] [gowalla] β=0.65 → K=5 R=1
+[01:56:20] [gowalla] β=0.70 → K=5 R=1
+[01:56:31] [gowalla] β=0.7311 → K=5 R=1
+[01:56:41] [gowalla] β=0.75 → K=5 R=1
+[01:56:51] [gowalla] β=0.80 → K=4 R=1
+[01:57:01] [gowalla] β=0.85 → K=4 R=1
+[01:57:11] [gowalla] β=0.90 → K=4 R=1
+[01:57:21] [gowalla] === DONE ===
+[01:57:21] [lastfm_1k] ws=309MB bufsz=14 l0=358
+[01:57:21] [lastfm_1k] reboot VM ssd=6144 bufsz=14MB
+/dev/dax0.0
+[01:58:48] [lastfm_1k] β=0.50 → K=31 R=1
+[01:59:02] [lastfm_1k] β=0.55 → K=29 R=1
+[01:59:15] [lastfm_1k] β=0.60 → K=27 R=1
+[01:59:29] [lastfm_1k] β=0.65 → K=25 R=1
+[01:59:43] [lastfm_1k] β=0.70 → K=23 R=1
+[01:59:57] [lastfm_1k] β=0.7311 → K=22 R=1
+[02:00:10] [lastfm_1k] β=0.75 → K=22 R=1
+[02:00:24] [lastfm_1k] β=0.80 → K=21 R=1
+[02:00:37] [lastfm_1k] β=0.85 → K=19 R=1
+[02:00:51] [lastfm_1k] β=0.90 → K=18 R=1
+[02:01:04] [lastfm_1k] === DONE ===
+[02:01:04] [taobao] ws=102MB bufsz=4 l0=102
+[02:01:04] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[02:02:25] [taobao] β=0.50 → K=38 R=1
+[02:02:33] [taobao] β=0.55 → K=37 R=1
+[02:02:40] [taobao] β=0.60 → K=36 R=1
+[02:02:48] [taobao] β=0.65 → K=35 R=1
+[02:02:55] [taobao] β=0.70 → K=34 R=1
+[02:03:03] [taobao] β=0.7311 → K=33 R=1
+[02:03:10] [taobao] β=0.75 → K=33 R=1
+[02:03:18] [taobao] β=0.80 → K=32 R=1
+[02:03:26] [taobao] β=0.85 → K=31 R=1
+[02:03:33] [taobao] β=0.90 → K=30 R=1
+[02:03:40] [taobao] === DONE ===
+[02:03:40] [bookcrossing] ws=81MB bufsz=3 l0=76
+[02:03:40] [bookcrossing] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[02:05:02] [bookcrossing] β=0.50 → K=13 R=1
+[02:05:07] [bookcrossing] β=0.55 → K=12 R=1
+[02:05:12] [bookcrossing] β=0.60 → K=11 R=1
+[02:05:16] [bookcrossing] β=0.65 → K=11 R=1
+[02:05:21] [bookcrossing] β=0.70 → K=10 R=1
+[02:05:26] [bookcrossing] β=0.7311 → K=10 R=1
+[02:05:31] [bookcrossing] β=0.75 → K=10 R=1
+[02:05:37] [bookcrossing] β=0.80 → K=9 R=1
+[02:05:42] [bookcrossing] β=0.85 → K=9 R=1
+[02:05:47] [bookcrossing] β=0.90 → K=9 R=1
+[02:05:52] [bookcrossing] === DONE ===
+[02:05:52] [lastfm_360k] ws=67MB bufsz=3 l0=76
+[02:05:52] [lastfm_360k] reboot VM ssd=6144 bufsz=3MB
+/dev/dax0.0
+[02:07:18] [lastfm_360k] β=0.50 → K=9 R=1
+[02:07:28] [lastfm_360k] β=0.55 → K=9 R=1
+[02:07:37] [lastfm_360k] β=0.60 → K=9 R=1
+[02:07:47] [lastfm_360k] β=0.65 → K=8 R=1
+[02:07:56] [lastfm_360k] β=0.70 → K=8 R=1
+[02:08:05] [lastfm_360k] β=0.7311 → K=8 R=1
+[02:08:15] [lastfm_360k] β=0.75 → K=8 R=1
+[02:08:24] [lastfm_360k] β=0.80 → K=8 R=1
+[02:08:33] [lastfm_360k] β=0.85 → K=8 R=1
+[02:08:42] [lastfm_360k] β=0.90 → K=8 R=1
+[02:08:51] [lastfm_360k] === DONE ===
+[02:08:51] [amazon_Electronics] ws=63MB bufsz=2 l0=51
+[02:08:51] [amazon_Electronics] reboot VM ssd=6144 bufsz=2MB
+/dev/dax0.0
+[02:10:17] [amazon_Electronics] β=0.50 → K=14 R=1
+[02:10:22] [amazon_Electronics] β=0.55 → K=13 R=1
+[02:10:27] [amazon_Electronics] β=0.60 → K=13 R=1
+[02:10:32] [amazon_Electronics] β=0.65 → K=13 R=1
+[02:10:37] [amazon_Electronics] β=0.70 → K=13 R=1
+[02:10:42] [amazon_Electronics] β=0.7311 → K=12 R=1
+[02:10:47] [amazon_Electronics] β=0.75 → K=12 R=1
+[02:10:51] [amazon_Electronics] β=0.80 → K=12 R=1
+[02:10:56] [amazon_Electronics] β=0.85 → K=11 R=1
+[02:11:00] [amazon_Electronics] β=0.90 → K=11 R=1
+[02:11:05] [amazon_Electronics] === DONE ===
+[02:11:05] [yelp] ws=37MB bufsz=1 l0=25
+[02:11:05] [yelp] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[02:12:26] [yelp] β=0.50 → K=6 R=1
+[02:12:31] [yelp] β=0.55 → K=6 R=1
+[02:12:36] [yelp] β=0.60 → K=6 R=1
+[02:12:41] [yelp] β=0.65 → K=6 R=1
+[02:12:45] [yelp] β=0.70 → K=5 R=1
+[02:12:49] [yelp] β=0.7311 → K=5 R=1
+[02:12:53] [yelp] β=0.75 → K=5 R=1
+[02:12:58] [yelp] β=0.80 → K=5 R=1
+[02:13:03] [yelp] β=0.85 → K=5 R=1
+[02:13:07] [yelp] β=0.90 → K=5 R=1
+[02:13:11] [yelp] === DONE ===
+[02:13:11] [retailrocket] ws=36MB bufsz=1 l0=25
+[02:13:11] [retailrocket] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[02:14:33] [retailrocket] β=0.50 → K=14 R=2
+[02:14:39] [retailrocket] β=0.55 → K=13 R=2
+[02:14:44] [retailrocket] β=0.60 → K=13 R=2
+[02:14:50] [retailrocket] β=0.65 → K=13 R=1
+[02:14:56] [retailrocket] β=0.70 → K=12 R=1
+[02:15:02] [retailrocket] β=0.7311 → K=12 R=1
+[02:15:08] [retailrocket] β=0.75 → K=12 R=1
+[02:15:14] [retailrocket] β=0.80 → K=12 R=1
+[02:15:20] [retailrocket] β=0.85 → K=12 R=1
+[02:15:26] [retailrocket] β=0.90 → K=11 R=1
+[02:15:31] [retailrocket] === DONE ===
+[02:15:31] [foursquare] ws=25MB bufsz=1 l0=25
+[02:15:31] [foursquare] reboot VM ssd=6144 bufsz=1MB
+/dev/dax0.0
+[02:17:03] [foursquare] β=0.50 → K=11 R=1
+[02:17:14] [foursquare] β=0.55 → K=10 R=1
+[02:17:25] [foursquare] β=0.60 → K=10 R=1
+[02:17:35] [foursquare] β=0.65 → K=10 R=1
+[02:17:46] [foursquare] β=0.70 → K=10 R=1
+[02:17:57] [foursquare] β=0.7311 → K=9 R=1
+[02:18:08] [foursquare] β=0.75 → K=9 R=1
+[02:18:18] [foursquare] β=0.80 → K=9 R=1
+[02:18:29] [foursquare] β=0.85 → K=9 R=1
+[02:18:39] [foursquare] β=0.90 → K=9 R=1
+[02:18:50] [foursquare] === DONE ===
+[02:18:50] === ALL DONE ===
+[02:18:50] Step 5: EWMA hyperparam sweep...
+[02:18:50] === START EWMA hyperparam sweep ===
+[02:18:50] [taobao] ws=102MB bufsz=4 l0=102
+[02:18:50] [taobao] reboot VM ssd=6144 bufsz=4MB
+/dev/dax0.0
+[02:20:16] [taobao] α=0.05 warmup=0
+[02:20:19] [taobao] α=0.05 warmup=50
+[02:20:22] [taobao] α=0.05 warmup=100
+[02:20:25] [taobao] α=0.05 warmup=200
+[02:20:27] [taobao] α=0.05 warmup=500
+[02:20:30] [taobao] α=0.10 warmup=0
+[02:20:33] [taobao] α=0.10 warmup=50
+[02:20:35] [taobao] α=0.10 warmup=100
+[02:20:38] [taobao] α=0.10 warmup=200
+[02:20:41] [taobao] α=0.10 warmup=500
+[02:20:44] [taobao] α=0.20 warmup=0
+[02:20:47] [taobao] α=0.20 warmup=50
+[02:20:50] [taobao] α=0.20 warmup=100
+[02:20:53] [taobao] α=0.20 warmup=200
+[02:20:56] [taobao] α=0.20 warmup=500
+[02:20:59] [taobao] α=0.30 warmup=0
+[02:21:01] [taobao] α=0.30 warmup=50
+[02:21:05] [taobao] α=0.30 warmup=100
+[02:21:07] [taobao] α=0.30 warmup=200
+[02:21:10] [taobao] α=0.30 warmup=500
+[02:21:13] [taobao] α=0.50 warmup=0
+[02:21:16] [taobao] α=0.50 warmup=50
+[02:21:18] [taobao] α=0.50 warmup=100
+[02:21:21] [taobao] α=0.50 warmup=200
+[02:21:24] [taobao] α=0.50 warmup=500
+[02:21:27] [taobao] === DONE ===
+[02:21:27] [criteo_kaggle] ws=5380MB bufsz=252 l0=6451
+[02:21:27] [criteo_kaggle] reboot VM ssd=6144 bufsz=252MB
+/dev/dax0.0
+[02:22:53] [criteo_kaggle] α=0.05 warmup=0
+[02:23:15] [criteo_kaggle] α=0.05 warmup=50
+[02:23:36] [criteo_kaggle] α=0.05 warmup=100
+[02:23:57] [criteo_kaggle] α=0.05 warmup=200
+[02:24:19] [criteo_kaggle] α=0.05 warmup=500
+[02:24:43] [criteo_kaggle] α=0.10 warmup=0
+[02:25:03] [criteo_kaggle] α=0.10 warmup=50
+[02:25:23] [criteo_kaggle] α=0.10 warmup=100
+[02:25:45] [criteo_kaggle] α=0.10 warmup=200
+[02:26:06] [criteo_kaggle] α=0.10 warmup=500
+[02:26:30] [criteo_kaggle] α=0.20 warmup=0
+[02:26:51] [criteo_kaggle] α=0.20 warmup=50
+[02:27:13] [criteo_kaggle] α=0.20 warmup=100
+[02:27:33] [criteo_kaggle] α=0.20 warmup=200
+[02:27:54] [criteo_kaggle] α=0.20 warmup=500
+[02:28:19] [criteo_kaggle] α=0.30 warmup=0
+[02:28:39] [criteo_kaggle] α=0.30 warmup=50
+[02:29:00] [criteo_kaggle] α=0.30 warmup=100
+[02:29:22] [criteo_kaggle] α=0.30 warmup=200
+[02:29:43] [criteo_kaggle] α=0.30 warmup=500
+[02:30:07] [criteo_kaggle] α=0.50 warmup=0
+[02:30:27] [criteo_kaggle] α=0.50 warmup=50
+[02:30:48] [criteo_kaggle] α=0.50 warmup=100
+[02:31:12] [criteo_kaggle] α=0.50 warmup=200
+[02:31:34] [criteo_kaggle] α=0.50 warmup=500
+[02:31:57] [criteo_kaggle] === DONE ===
+[02:31:57] [avazu] ws=2307MB bufsz=108 l0=2764
+[02:31:57] [avazu] reboot VM ssd=6144 bufsz=108MB
+/dev/dax0.0
+[02:33:23] [avazu] α=0.05 warmup=0
+[02:33:27] [avazu] α=0.05 warmup=50
+[02:33:30] [avazu] α=0.05 warmup=100
+[02:33:34] [avazu] α=0.05 warmup=200
+[02:33:38] [avazu] α=0.05 warmup=500
+[02:33:41] [avazu] α=0.10 warmup=0
+[02:33:45] [avazu] α=0.10 warmup=50
+[02:33:48] [avazu] α=0.10 warmup=100
+[02:33:52] [avazu] α=0.10 warmup=200
+[02:33:55] [avazu] α=0.10 warmup=500
+[02:33:59] [avazu] α=0.20 warmup=0
+[02:34:02] [avazu] α=0.20 warmup=50
+[02:34:06] [avazu] α=0.20 warmup=100
+[02:34:10] [avazu] α=0.20 warmup=200
+[02:34:14] [avazu] α=0.20 warmup=500
+[02:34:17] [avazu] α=0.30 warmup=0
+[02:34:21] [avazu] α=0.30 warmup=50
+[02:34:24] [avazu] α=0.30 warmup=100
+[02:34:28] [avazu] α=0.30 warmup=200
+[02:34:31] [avazu] α=0.30 warmup=500
+[02:34:35] [avazu] α=0.50 warmup=0
+[02:34:38] [avazu] α=0.50 warmup=50
+[02:34:42] [avazu] α=0.50 warmup=100
+[02:34:45] [avazu] α=0.50 warmup=200
+[02:34:49] [avazu] α=0.50 warmup=500
+[02:34:52] [avazu] === DONE ===
+[02:34:52] === ALL DONE ===
+[02:34:53] === Route B orchestrator ALL DONE ===

route_b_watchdog.log

@@ -0,0 +1,9 @@
+[01:22:35] watchdog start, waiting for orchestrator ALL_DONE...
+[02:35:35] orchestrator done, launching taobao_n17...
+[02:35:35] === START taobao n=17 ===
+/dev/dax0.0
+[02:37:02] mode=9
+[02:37:24] mode=11
+[02:37:44] mode=13
+[02:38:05] === DONE ===
+[02:38:05] watchdog done.

taobao_n17/run_20260501_023535.log

@@ -0,0 +1,6 @@
+[02:35:35] === START taobao n=17 ===
+/dev/dax0.0
+[02:37:02] mode=9
+[02:37:24] mode=11
+[02:37:44] mode=13
+[02:38:05] === DONE ===