fix: e2e aiperf profiling on NAT dataset (#5990)

Signed-off-by: PeaBrane <yanrpei@gmail.com>

benchmarks/nat_trace/convert.py

@@ -19,7 +19,7 @@ import re
 from collections import defaultdict
 
 from aiperf.common.tokenizer import Tokenizer
-from aiperf.dataset.synthesis.rolling_hasher import texts_to_hashes
+from aiperf.dataset.synthesis.rolling_hasher import RollingHasher
 from tqdm import tqdm
 
 
@@ -49,8 +49,8 @@ def parse_args():
     parser.add_argument(
         "--block-size",
         type=int,
-        default=128,
-        help="Block size for hash generation (default: 128)",
+        default=64,
+        help="Block size for hash generation (default: 64)",
     )
     parser.add_argument(
         "--num-requests",
@@ -64,6 +64,12 @@ def parse_args():
         default=0,
         help="Skip the first N requests (default: 0)",
     )
+    parser.add_argument(
+        "--delay",
+        type=int,
+        default=500,
+        help="Delay in ms to add between LLM calls within a session (simulates tool call timing). If not specified, no delay field is added.",
+    )
     return parser.parse_args()
 
 
@@ -173,6 +179,39 @@ def extract_llm_calls(request: dict) -> list[dict]:
     return llm_calls
 
 
+def texts_to_hashes_and_lengths(
+    tokenizer: Tokenizer,
+    texts: list[str],
+    block_size: int,
+) -> tuple[list[list[int]], list[int]]:
+    """
+    Convert texts to hash IDs and token lengths.
+
+    Returns:
+        Tuple of (hash_ids_list, token_lengths) where:
+        - hash_ids_list: List of hash ID sequences, one per input text
+        - token_lengths: List of token counts, one per input text
+    """
+    hasher = RollingHasher(block_size=block_size)
+    hash_results: list[list[int]] = []
+    length_results: list[int] = []
+
+    for text in texts:
+        tokens = tokenizer.encode(text)
+        length_results.append(len(tokens))
+
+        blocks: list[list[int]] = [
+            tokens[i : i + block_size] for i in range(0, len(tokens), block_size)
+        ]
+        if blocks:
+            hashes = hasher.hash_token_blocks(blocks)
+            hash_results.append(hashes)
+        else:
+            hash_results.append([])
+
+    return hash_results, length_results
+
+
 def chat_inputs_to_text(chat_inputs: list) -> str:
     """
     Convert chat_inputs array to a single text string for hashing.
@@ -200,6 +239,7 @@ def convert_to_mooncake(
     block_size: int,
     skip_requests: int = 0,
     num_requests: int | None = None,
+    delay: int | None = None,
 ) -> list[dict]:
     """
     Convert NAT requests to mooncake format.
@@ -210,6 +250,7 @@ def convert_to_mooncake(
         block_size: Block size for hash generation
         skip_requests: Number of requests to skip
         num_requests: Maximum number of requests to process
+        delay: Delay in ms to add between LLM calls within a session
 
     Returns:
         List of mooncake-format dicts
@@ -222,7 +263,7 @@ def convert_to_mooncake(
     print(f"Processing {len(requests)} requests...")
 
     # Phase 1: Collect all texts and metadata
-    all_entries = []  # List of (session_id, prompt_tokens, completion_tokens, text)
+    all_entries = []  # List of (session_id, completion_tokens, text)
 
     for req in tqdm(requests, desc="Extracting LLM calls"):
         request_number = req.get("request_number", 0)
@@ -235,10 +276,6 @@ def convert_to_mooncake(
             continue
 
         for call in llm_calls:
-            # Skip calls with zero tokens (might be incomplete)
-            if call["prompt_tokens"] == 0:
-                continue
-
             # Convert chat_inputs to text for hashing
             text = chat_inputs_to_text(call["chat_inputs"])
 
@@ -251,7 +288,6 @@ def convert_to_mooncake(
             all_entries.append(
                 (
                     session_id,
-                    call["prompt_tokens"],
                     call["completion_tokens"],
                     text,
                 )
@@ -261,24 +297,33 @@ def convert_to_mooncake(
         print("No valid LLM calls found")
         return []
 
-    # Phase 2: Batch hash all texts at once (single hasher instance)
-    all_texts = [entry[3] for entry in all_entries]
-    print(f"Hashing {len(all_texts)} texts...")
+    # Phase 2: Tokenize texts to get hash IDs and token lengths
+    all_texts = [entry[2] for entry in all_entries]
+    print(f"Tokenizing and hashing {len(all_texts)} texts...")
 
     tokenizer = Tokenizer.from_pretrained(tokenizer_name)
-    all_hash_ids = texts_to_hashes(tokenizer, all_texts, block_size)
+    all_hash_ids, all_input_lengths = texts_to_hashes_and_lengths(
+        tokenizer, all_texts, block_size
+    )
 
     # Phase 3: Build mooncake entries
     mooncake_data = []
-    for (session_id, prompt_tokens, completion_tokens, _), hash_ids in zip(
-        all_entries, all_hash_ids, strict=True
+    seen_sessions = set()
+    for (session_id, completion_tokens, _), hash_ids, input_length in zip(
+        all_entries, all_hash_ids, all_input_lengths, strict=True
     ):
         mooncake_entry = {
             "session_id": session_id,
-            "input_length": prompt_tokens,
+            "input_length": input_length,
             "output_length": completion_tokens,
             "hash_ids": hash_ids,
         }
+        # Add delay for all but the first entry in each session
+        if delay is not None:
+            if session_id in seen_sessions:
+                mooncake_entry["delay"] = delay
+            else:
+                seen_sessions.add(session_id)
         mooncake_data.append(mooncake_entry)
 
     return mooncake_data
@@ -386,6 +431,7 @@ def main():
         args.block_size,
         skip_requests=args.skip_requests,
         num_requests=args.num_requests,
+        delay=args.delay,
     )
 
     # Print statistics

benchmarks/pyproject.toml

@@ -41,7 +41,7 @@ classifiers = [
 
 dependencies = [
     "aiconfigurator[webapp] @ git+https://github.com/ai-dynamo/aiconfigurator.git@release/0.6.0",
-    "aiperf @ git+https://github.com/ai-dynamo/aiperf.git@40530ed231fb01ae1cfe3c9d22e43a0e7143780b",
+    "aiperf @ git+https://github.com/ai-dynamo/aiperf.git@c3fc969e9e30e9ddad35b2f613aa7c1d418f2de9",
     "matplotlib",
     "networkx",
     "numpy",

benchmarks/router/README.md

@@ -41,7 +41,8 @@ This will start both etcd and NATS with the required configurations in the backg
 - **`ping.sh`** - Simple test script to verify the setup is working
 - **`prefix_ratio_benchmark.py`** - Main benchmarking script that sweeps prefix ratios
 - **`real_data_benchmark.py`** - Benchmarking script that uses real mooncake-style trace data
-- **`plot_prefix_ratio_comparison.py`** - Generates comparison plots from benchmark results
+- **`agent_benchmark.py`** - Concurrency-based benchmarking for multi-turn conversation traces
+- **`mock_server.py`** - Simple mock server to receive and log requests from aiperf
 
 ## Usage Instructions
 
@@ -245,6 +246,61 @@ python real_data_benchmark.py --input-dataset trace.jsonl --prefix-root-multipli
 > ```
 > However, by the time of release, the aiperf version included in the vLLM runtime container should be up to date enough to use as-is.
 
+### Step 4 (Alternative): Agent Benchmark (Concurrency-Based Multi-Turn)
+
+For benchmarking with multi-turn conversation traces using concurrency-based load generation (instead of timestamp-based replay), use `agent_benchmark.py`. This is useful for testing how the system handles multiple concurrent agent sessions.
+
+```bash
+python agent_benchmark.py --input-dataset trace.jsonl --concurrency 10
+```
+
+**Key parameters:**
+- `--concurrency`: Number of concurrent sessions to maintain (default: 10)
+- `--delay`: Override delay (ms) between turns within a session. Set to 0 to remove all delays.
+
+Examples:
+
+```bash
+# Run with 20 concurrent sessions using delays from trace file
+python agent_benchmark.py --input-dataset trace.jsonl --concurrency 20
+
+# Run with no delays between turns (stress test)
+python agent_benchmark.py --input-dataset trace.jsonl --concurrency 10 --delay 0
+
+# Run with fixed 1-second delay between turns
+python agent_benchmark.py --input-dataset trace.jsonl --concurrency 10 --delay 1000
+```
+
+### Trace Dataset Format (JSONL)
+
+Both `real_data_benchmark.py` and `agent_benchmark.py` accept trace datasets in JSONL format (one JSON object per line). The format is compatible with [Mooncake trace format](https://github.com/kvcache-ai/Mooncake).
+
+#### Fields
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `input_length` | int | Number of input tokens for this request |
+| `output_length` | int | Number of output tokens to generate |
+| `session_id` | string | Groups turns into multi-turn conversations. Requests with the same `session_id` are processed sequentially. |
+| `hash_ids` | list[int] | List of hash IDs representing prefix blocks for KV cache routing. Shared hash IDs indicate shared prefixes. |
+| `delay` | int | Delay in milliseconds to wait before sending this turn (applied after the previous turn in the same session completes). Not applied to first turns. |
+
+#### Example Trace File
+
+```jsonl
+{"session_id": "conv_0", "input_length": 9176, "output_length": 152, "hash_ids": [0, 1, 2, 3, 4, 5]}
+{"session_id": "conv_0", "input_length": 9368, "output_length": 104, "hash_ids": [0, 1, 2, 3, 4, 5, 6, 7], "delay": 500}
+{"session_id": "conv_0", "input_length": 9516, "output_length": 164, "hash_ids": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9], "delay": 500}
+{"session_id": "conv_1", "input_length": 9445, "output_length": 143, "hash_ids": [0, 1, 2, 10, 11, 12, 13]}
+{"session_id": "conv_1", "input_length": 9628, "output_length": 123, "hash_ids": [0, 1, 2, 10, 11, 12, 13, 14, 15], "delay": 500}
+```
+
+In this example:
+- `conv_0` and `conv_1` are two separate conversations that can run concurrently
+- Within each conversation, turns are processed sequentially
+- Subsequent turns have a 500ms delay after the previous turn completes
+- `hash_ids` show prefix sharing: both conversations share prefix blocks `[0, 1, 2]`
+
 ## Benchmarking Results
 
 We benchmarked the Dynamo KV Router against a baseline round-robin routing strategy to evaluate the performance benefits of cache-aware routing. The experiments were conducted using deepseek-ai/DeepSeek-R1-Distill-Llama-8B on 8 L40S GPUs under aggregated serving, with the following configuration:

benchmarks/router/agent_benchmark.py

+#!/usr/bin/env python3
+
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+"""
+Agent benchmark script for running concurrency-based benchmarks with multi-turn
+conversation traces. Uses aiperf with concurrency mode to run multiple sessions
+in parallel while maintaining sequential ordering within each session.
+
+Expected input JSON format (JSONL - one JSON object per line):
+{
+    "session_id": "conv_0",           # Groups turns into conversations (required for multi-turn)
+    "input_length": 9176,             # Number of input tokens (required)
+    "output_length": 152,             # Number of output tokens (required)
+    "hash_ids": [0, 1, 2, ...],       # List of hash IDs for prefix caching (optional)
+    "delay": 500                      # Delay in ms before this turn (optional, applied after previous turn completes)
+}
+"""
+
+import argparse
+import json
+import os
+import subprocess
+
+from common import (
+    DEFAULT_BLOCK_SIZE,
+    add_common_args,
+    get_common_aiperf_flags,
+    resolve_tokenizer,
+    setup_logger,
+)
+
+logger = setup_logger(__name__)
+
+
+def count_dataset_entries(input_dataset):
+    """Count the number of entries in a JSONL dataset file."""
+    count = 0
+    with open(input_dataset, "r") as f:
+        for line in f:
+            if line.strip():
+                count += 1
+    return count
+
+
+def get_aiperf_cmd(
+    model,
+    tokenizer,
+    input_dataset,
+    artifact_dir,
+    seed,
+    concurrency,
+    block_size,
+    request_count,
+    url="http://localhost:8000",
+):
+    """Build aiperf command for concurrency-based trace benchmarking."""
+    cmd = [
+        "aiperf",
+        "profile",
+        "--model",
+        model,
+        "--tokenizer",
+        tokenizer,
+        "--url",
+        url,
+        "--input-file",
+        input_dataset,
+        "--custom-dataset-type",
+        "mooncake_trace",
+        "--concurrency",
+        str(concurrency),
+        "--request-count",
+        str(request_count),
+        "--prompt-input-tokens-block-size",
+        str(block_size),
+        "--random-seed",
+        str(seed),
+        "--artifact-dir",
+        artifact_dir,
+    ]
+    cmd.extend(get_common_aiperf_flags())
+    return cmd
+
+
+def prepare_dataset(input_dataset, output_path, delay_override=None):
+    """
+    Prepare the dataset, optionally overriding delay values.
+
+    Args:
+        input_dataset: Path to input JSONL file
+        output_path: Path to write modified dataset
+        delay_override: If set, override all delay values with this value (in ms).
+                       Use 0 to remove delays entirely.
+
+    Returns:
+        Path to the dataset to use (original or modified)
+    """
+    if delay_override is None:
+        return input_dataset
+
+    logger.info(f"Overriding delay values with: {delay_override}ms")
+
+    requests = []
+    with open(input_dataset, "r") as f:
+        for line in f:
+            line = line.strip()
+            if line:
+                requests.append(json.loads(line))
+
+    # Track sessions to know which entries are first turns (no delay on first turn)
+    session_first_seen = set()
+
+    for request in requests:
+        session_id = request.get("session_id")
+
+        if session_id is not None and session_id not in session_first_seen:
+            # First turn of a session - remove delay if present
+            session_first_seen.add(session_id)
+            request.pop("delay", None)
+        elif delay_override == 0:
+            # Remove delay entirely
+            request.pop("delay", None)
+        else:
+            # Override delay for subsequent turns
+            request["delay"] = delay_override
+
+    with open(output_path, "w") as f:
+        for request in requests:
+            f.write(json.dumps(request) + "\n")
+
+    logger.info(f"Modified dataset saved to: {output_path}")
+    return output_path
+
+
+def run_benchmark(
+    model,
+    tokenizer,
+    trace_dataset,
+    artifact_dir,
+    url,
+    seed,
+    concurrency,
+    block_size,
+    request_count,
+):
+    """Run aiperf benchmark with concurrency mode."""
+    aiperf_cmd = get_aiperf_cmd(
+        model,
+        tokenizer,
+        trace_dataset,
+        artifact_dir,
+        seed,
+        concurrency,
+        block_size,
+        request_count,
+        url,
+    )
+
+    logger.info(
+        f"Running aiperf with concurrency={concurrency}, request_count={request_count}"
+    )
+    logger.info(f"Dataset: {trace_dataset}")
+    logger.info(f"Command: {' '.join(aiperf_cmd)}")
+
+    try:
+        subprocess.run(aiperf_cmd, check=True)
+        logger.info("AIPerf profiling completed successfully")
+    except subprocess.CalledProcessError as e:
+        logger.error(f"AIPerf failed with error code: {e.returncode}")
+        raise
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Run concurrency-based benchmark with multi-turn conversation traces"
+    )
+
+    # Common arguments
+    add_common_args(parser)
+
+    parser.add_argument(
+        "--output-dir",
+        type=str,
+        default="agent_benchmark_results",
+        help="Output directory for results",
+    )
+
+    # Dataset configuration
+    parser.add_argument(
+        "--input-dataset",
+        type=str,
+        required=True,
+        help="Path to the input trace dataset file (JSONL format)",
+    )
+
+    # Benchmark configuration
+    parser.add_argument(
+        "--concurrency",
+        type=int,
+        default=10,
+        help="Number of concurrent sessions to maintain (default: 10)",
+    )
+    parser.add_argument(
+        "--delay",
+        type=int,
+        default=None,
+        help="Override delay (ms) between turns within a session. "
+        "If not set, uses delay values from the trace file. "
+        "Set to 0 to remove all delays.",
+    )
+    parser.add_argument(
+        "--block-size",
+        type=int,
+        default=DEFAULT_BLOCK_SIZE,
+        help=f"Block size for prompt generation from hash_ids (default: {DEFAULT_BLOCK_SIZE})",
+    )
+    parser.add_argument(
+        "--request-count",
+        type=int,
+        default=None,
+        help="Number of requests to send. If not set, defaults to number of entries in input dataset.",
+    )
+
+    args = parser.parse_args()
+    resolve_tokenizer(args)
+
+    # Default request_count to dataset entry count if not specified
+    if args.request_count is None:
+        args.request_count = count_dataset_entries(args.input_dataset)
+        logger.info(
+            f"Request count not specified, using dataset entry count: {args.request_count}"
+        )
+
+    # Create output directory
+    os.makedirs(args.output_dir, exist_ok=True)
+
+    # Prepare dataset (apply delay override if specified)
+    if args.delay is not None:
+        modified_dataset_path = os.path.join(args.output_dir, "modified_trace.jsonl")
+        trace_dataset_path = prepare_dataset(
+            args.input_dataset, modified_dataset_path, args.delay
+        )
+    else:
+        trace_dataset_path = args.input_dataset
+        logger.info(f"Using original trace dataset: {trace_dataset_path}")
+
+    # Run benchmark
+    artifact_dir = os.path.join(args.output_dir, "aiperf_artifacts")
+    os.makedirs(artifact_dir, exist_ok=True)
+
+    run_benchmark(
+        args.model,
+        args.tokenizer,
+        trace_dataset_path,
+        artifact_dir,
+        args.url,
+        args.seed,
+        args.concurrency,
+        args.block_size,
+        args.request_count,
+    )
+
+    logger.info(f"Results saved to: {artifact_dir}")
+
+
+if __name__ == "__main__":
+    main()

benchmarks/router/common.py

+#!/usr/bin/env python3
+
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+"""Common utilities shared across router benchmark scripts."""
+
+import logging
+
+# Default values
+DEFAULT_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
+DEFAULT_URL = "http://localhost:8000"
+DEFAULT_SEED = 0
+DEFAULT_BLOCK_SIZE = 64
+DEFAULT_MOONCAKE_BLOCK_SIZE = 512
+
+
+def setup_logger(name: str) -> logging.Logger:
+    """Setup and return a logger with standard formatting."""
+    logger = logging.getLogger(name)
+    logger.setLevel(logging.INFO)
+    console_handler = logging.StreamHandler()
+    console_handler.setLevel(logging.INFO)
+    formatter = logging.Formatter(
+        "%(asctime)s - %(name)s - %(levelname)s - %(message)s", "%Y-%m-%d %H:%M:%S"
+    )
+    console_handler.setFormatter(formatter)
+    logger.addHandler(console_handler)
+    return logger
+
+
+def add_common_args(parser):
+    """Add common CLI arguments shared across benchmark scripts."""
+    parser.add_argument(
+        "--model",
+        type=str,
+        default=DEFAULT_MODEL,
+        help="Model name",
+    )
+    parser.add_argument(
+        "--tokenizer",
+        type=str,
+        default=None,
+        help="Tokenizer name (defaults to model)",
+    )
+    parser.add_argument(
+        "--url",
+        type=str,
+        default=DEFAULT_URL,
+        help="Server URL",
+    )
+    parser.add_argument(
+        "--seed",
+        type=int,
+        default=DEFAULT_SEED,
+        help="Random seed for reproducibility (default: 0)",
+    )
+    parser.add_argument(
+        "--use-expected-osl",
+        action="store_true",
+        help="Pass expected_output_tokens to nvext for router tracking",
+    )
+
+
+def resolve_tokenizer(args):
+    """Set tokenizer to model if not specified."""
+    if args.tokenizer is None:
+        args.tokenizer = args.model
+
+
+def get_common_aiperf_flags():
+    """Return common aiperf flags used across benchmarks."""
+    return [
+        "--endpoint-type",
+        "chat",
+        "--endpoint",
+        "v1/chat/completions",
+        "--streaming",
+        "--extra-inputs",
+        "ignore_eos:true",
+        "--no-gpu-telemetry",
+        "-H",
+        "Authorization: Bearer NOT USED",
+        "-H",
+        "Accept: text/event-stream",
+    ]

benchmarks/router/mock_server.py

+#!/usr/bin/env python3
+
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+"""
+Simple mock server to receive and log requests from aiperf loadgen.
+
+Prints each request with timestamp, useful for debugging request patterns,
+verifying session ordering, and inspecting delays between requests.
+
+Usage:
+    python mock_server.py --port 8000
+
+Then run aiperf pointing to this server:
+    python agent_benchmark.py --input-dataset trace.jsonl --url http://localhost:8000
+"""
+
+import argparse
+import asyncio
+import json
+import time
+from datetime import datetime
+
+from aiohttp import web
+
+
+class RequestLogger:
+    """Tracks and logs incoming requests with timing information."""
+
+    def __init__(self, output_file: str | None = None, payload_file: str | None = None):
+        self.request_count = 0
+        self.start_time = None
+        self.session_last_seen: dict[str, float] = {}
+        self.output_file = output_file
+        self.payload_file = payload_file
+        self._file_handle = None
+        self._payload_handle = None
+
+        if output_file:
+            self._file_handle = open(output_file, "w")
+        if payload_file:
+            self._payload_handle = open(payload_file, "w")
+
+    def close(self):
+        """Close the output files if open."""
+        if self._file_handle:
+            self._file_handle.close()
+        if self._payload_handle:
+            self._payload_handle.close()
+
+    def log_request(self, request_data: dict, headers: dict) -> None:
+        """Log a request with timestamp and relevant metadata."""
+        now = time.time()
+        if self.start_time is None:
+            self.start_time = now
+
+        self.request_count += 1
+        elapsed = now - self.start_time
+        timestamp = datetime.now().strftime("%H:%M:%S.%f")[:-3]
+
+        # Extract useful fields
+        model = request_data.get("model", "unknown")
+        messages = request_data.get("messages", [])
+        # Support both max_completion_tokens (new) and max_tokens (legacy)
+        max_tokens = request_data.get(
+            "max_completion_tokens", request_data.get("max_tokens", "N/A")
+        )
+
+        # Check for session tracking via x-correlation-id header
+        correlation_id = headers.get(
+            "x-correlation-id", headers.get("X-Correlation-Id")
+        )
+
+        # Calculate delay since last request from same session
+        session_delay = None
+        if correlation_id and correlation_id in self.session_last_seen:
+            session_delay = (now - self.session_last_seen[correlation_id]) * 1000  # ms
+        if correlation_id:
+            self.session_last_seen[correlation_id] = now
+
+        # Build log line
+        log_parts = [
+            f"[{timestamp}]",
+            f"#{self.request_count:4d}",
+            f"elapsed={elapsed:7.2f}s",
+            f"model={model}",
+            f"msgs={len(messages)}",
+            f"max_tokens={max_tokens}",
+        ]
+
+        if correlation_id:
+            # Truncate for display
+            short_id = correlation_id[:8] if len(correlation_id) > 8 else correlation_id
+            log_parts.append(f"session={short_id}")
+
+        if session_delay is not None:
+            log_parts.append(f"delay={session_delay:.0f}ms")
+
+        log_line = " | ".join(log_parts)
+        print(log_line)
+
+        if self._file_handle:
+            self._file_handle.write(log_line + "\n")
+            self._file_handle.flush()
+
+        # Dump full payload to separate file
+        if self._payload_handle:
+            payload_entry = {
+                "request_number": self.request_count,
+                "timestamp": timestamp,
+                "elapsed_s": elapsed,
+                "correlation_id": correlation_id,
+                "headers": {
+                    k: v
+                    for k, v in headers.items()
+                    if k.lower() not in ("authorization", "accept")
+                },
+                "body": request_data,
+            }
+            self._payload_handle.write(json.dumps(payload_entry) + "\n")
+            self._payload_handle.flush()
+
+
+logger: RequestLogger = None  # Initialized in main()
+
+
+async def health_check(request: web.Request) -> web.Response:
+    """Health check endpoint."""
+    return web.json_response({"status": "ok"})
+
+
+async def chat_completions(request: web.Request) -> web.StreamResponse:
+    """Handle /v1/chat/completions requests."""
+    try:
+        body = await request.json()
+    except json.JSONDecodeError:
+        return web.json_response({"error": "Invalid JSON"}, status=400)
+
+    # Log the request
+    headers = dict(request.headers)
+    logger.log_request(body, headers)
+
+    # Check if streaming is requested
+    stream = body.get("stream", False)
+    # Support both max_completion_tokens (new) and max_tokens (legacy)
+    max_tokens = body.get("max_completion_tokens", body.get("max_tokens", 10))
+
+    if stream:
+        # Return a streaming response
+        response = web.StreamResponse(
+            status=200,
+            headers={
+                "Content-Type": "text/event-stream",
+                "Cache-Control": "no-cache",
+                "Connection": "keep-alive",
+            },
+        )
+        await response.prepare(request)
+
+        # Send a minimal streaming response
+        # Each chunk simulates a token being generated
+        for i in range(max_tokens):
+            chunk = {
+                "id": f"chatcmpl-mock-{logger.request_count}",
+                "object": "chat.completion.chunk",
+                "created": int(time.time()),
+                "model": body.get("model", "mock"),
+                "choices": [
+                    {
+                        "index": 0,
+                        "delta": {"content": "x"}
+                        if i > 0
+                        else {"role": "assistant", "content": ""},
+                        "finish_reason": None,
+                    }
+                ],
+            }
+            await response.write(f"data: {json.dumps(chunk)}\n\n".encode())
+            await asyncio.sleep(0.001)  # Small delay between tokens
+
+        # Send final chunk with finish_reason
+        final_chunk = {
+            "id": f"chatcmpl-mock-{logger.request_count}",
+            "object": "chat.completion.chunk",
+            "created": int(time.time()),
+            "model": body.get("model", "mock"),
+            "choices": [
+                {
+                    "index": 0,
+                    "delta": {},
+                    "finish_reason": "stop",
+                }
+            ],
+        }
+        await response.write(f"data: {json.dumps(final_chunk)}\n\n".encode())
+        await response.write(b"data: [DONE]\n\n")
+        await response.write_eof()
+        return response
+    else:
+        # Non-streaming response
+        return web.json_response(
+            {
+                "id": f"chatcmpl-mock-{logger.request_count}",
+                "object": "chat.completion",
+                "created": int(time.time()),
+                "model": body.get("model", "mock"),
+                "choices": [
+                    {
+                        "index": 0,
+                        "message": {"role": "assistant", "content": "x" * max_tokens},
+                        "finish_reason": "stop",
+                    }
+                ],
+                "usage": {
+                    "prompt_tokens": 10,
+                    "completion_tokens": max_tokens,
+                    "total_tokens": 10 + max_tokens,
+                },
+            }
+        )
+
+
+async def models_list(request: web.Request) -> web.Response:
+    """Handle /v1/models endpoint."""
+    return web.json_response(
+        {
+            "object": "list",
+            "data": [
+                {
+                    "id": "mock-model",
+                    "object": "model",
+                    "created": int(time.time()),
+                    "owned_by": "mock",
+                }
+            ],
+        }
+    )
+
+
+def create_app() -> web.Application:
+    """Create the aiohttp application."""
+    app = web.Application()
+    app.router.add_get("/health", health_check)
+    app.router.add_get("/v1/models", models_list)
+    app.router.add_post("/v1/chat/completions", chat_completions)
+    return app
+
+
+def main():
+    global logger
+
+    parser = argparse.ArgumentParser(
+        description="Mock server to receive and log aiperf requests"
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=8000,
+        help="Port to listen on (default: 8000)",
+    )
+    parser.add_argument(
+        "--host",
+        type=str,
+        default="0.0.0.0",
+        help="Host to bind to (default: 0.0.0.0)",
+    )
+    parser.add_argument(
+        "--output",
+        type=str,
+        default="mock_server_logs.txt",
+        help="Output file for request logs (default: mock_server_logs.txt)",
+    )
+    parser.add_argument(
+        "--payload-output",
+        type=str,
+        default="mock_server_payloads.jsonl",
+        help="Output file for full request payloads in JSONL format (default: mock_server_payloads.jsonl)",
+    )
+    args = parser.parse_args()
+
+    # Initialize logger with output files
+    logger = RequestLogger(output_file=args.output, payload_file=args.payload_output)
+
+    print(f"Starting mock server on {args.host}:{args.port}")
+    print(f"Logging requests to: {args.output}")
+    print(f"Dumping payloads to: {args.payload_output}")
+    print("Waiting for requests from aiperf...")
+    print("-" * 80)
+
+    try:
+        app = create_app()
+        web.run_app(app, host=args.host, port=args.port, print=None)
+    finally:
+        logger.close()
+
+
+if __name__ == "__main__":
+    main()

benchmarks/router/ping.sh

@@ -16,6 +16,6 @@ curl -X POST http://localhost:${PORT}/v1/chat/completions \
         {"role": "user", "content": "What is 2+2?"}
     ],
     "stream": true,
-    "max_tokens": 10,
+    "max_completion_tokens": 10,
     "ignore_eos": true
     }'
\ No newline at end of file

benchmarks/router/prefix_ratio_benchmark.py

@@ -5,31 +5,27 @@
 
 import argparse
 import json
-import logging
 import os
 import subprocess
-from typing import Dict, List, Optional
+from typing import Dict, Optional
 
 import matplotlib
 
 matplotlib.use("Agg")  # Use non-interactive backend
 import matplotlib.pyplot as plt
-
-# Setup logging
-logger = logging.getLogger(__name__)
-logger.setLevel(logging.INFO)
-console_handler = logging.StreamHandler()
-console_handler.setLevel(logging.INFO)
-formatter = logging.Formatter(
-    "%(asctime)s - %(name)s - %(levelname)s - %(message)s", "%Y-%m-%d %H:%M:%S"
+from common import (
+    add_common_args,
+    get_common_aiperf_flags,
+    resolve_tokenizer,
+    setup_logger,
 )
-console_handler.setFormatter(formatter)
-logger.addHandler(console_handler)
+
+logger = setup_logger(__name__)
 
 
 def get_aiperf_cmd(
     model,
-    tokenizer,  # Add tokenizer parameter
+    tokenizer,
     prefix_ratio,
     isl,
     osl,
@@ -51,18 +47,13 @@ def get_aiperf_cmd(
         nvext_dict["expected_output_tokens"] = osl
     nvext_json = json.dumps({"nvext": nvext_dict})
 
-    return [
+    cmd = [
         "aiperf",
         "profile",
         "--model",
         model,
         "--tokenizer",
-        tokenizer,  # Use the tokenizer parameter instead of model
-        "--endpoint-type",
-        "chat",
-        "--endpoint",
-        "v1/chat/completions",
-        "--streaming",
+        tokenizer,
         "--url",
         url,
         "--synthetic-input-tokens-mean",
@@ -74,8 +65,6 @@ def get_aiperf_cmd(
         "--output-tokens-stddev",
         str(round(osl / 4)),
         "--extra-inputs",
-        "ignore_eos:true",
-        "--extra-inputs",
         nvext_json,
         "--concurrency",
         str(concurrency),
@@ -93,11 +82,9 @@ def get_aiperf_cmd(
         artifact_dir,
         "--dataset-sampling-strategy",
         "shuffle",
-        "-H",
-        "Authorization: Bearer NOT USED",
-        "-H",
-        "Accept: text/event-stream",
     ]
+    cmd.extend(get_common_aiperf_flags())
+    return cmd
 
 
 def get_aiperf_result(artifact_dir: str) -> dict:
@@ -117,90 +104,9 @@ def get_aiperf_result(artifact_dir: str) -> dict:
         return json.load(f)
 
 
-def run_benchmark_single_url(
-    model,
-    tokenizer,  # Add tokenizer parameter
-    prefix_ratio,
-    isl,
-    osl,
-    requests,
-    concurrency,
-    seed,
-    num_prefix_prompts,
-    artifact_dir,
-    url,
-    use_expected_osl=False,
-) -> Optional[Dict]:
-    """Run aiperf benchmark for a single URL"""
-    aiperf_cmd = get_aiperf_cmd(
-        model,
-        tokenizer,  # Pass tokenizer parameter
-        prefix_ratio,
-        isl,
-        osl,
-        requests,
-        concurrency,
-        seed,
-        num_prefix_prompts,
-        artifact_dir,
-        url,
-        use_expected_osl,
-    )
-
-    logger.info(f"Running command for URL {url}: {' '.join(aiperf_cmd)}")
-
-    try:
-        # Run aiperf and let it output directly to terminal
-        subprocess.run(aiperf_cmd, check=True)
-
-        logger.info(f"AIPerf profiling completed successfully for URL {url}")
-
-        aiperf_result = get_aiperf_result(artifact_dir)
-        return aiperf_result
-
-    except subprocess.CalledProcessError as e:
-        logger.error(f"AIPerf failed for URL {url} with error code: {e.returncode}")
-        return None
-
-
-def aggregate_results(results: List[Optional[Dict]]) -> Optional[Dict]:
-    """Aggregate results from multiple URLs"""
-    if not results:
-        return None
-
-    valid_results = [r for r in results if r is not None]
-    if not valid_results:
-        return None
-
-    # For TTFT percentiles, average across URLs
-    ttft_p25_values = [r["time_to_first_token"]["p25"] for r in valid_results]
-    ttft_p50_values = [r["time_to_first_token"]["p50"] for r in valid_results]
-    ttft_p75_values = [r["time_to_first_token"]["p75"] for r in valid_results]
-
-    # For ITL percentiles, average across URLs
-    itl_p25_values = [r["inter_token_latency"]["p25"] for r in valid_results]
-    itl_p50_values = [r["inter_token_latency"]["p50"] for r in valid_results]
-    itl_p75_values = [r["inter_token_latency"]["p75"] for r in valid_results]
-
-    aggregated = {
-        "time_to_first_token": {
-            "p25": sum(ttft_p25_values) / len(ttft_p25_values),
-            "p50": sum(ttft_p50_values) / len(ttft_p50_values),
-            "p75": sum(ttft_p75_values) / len(ttft_p75_values),
-        },
-        "inter_token_latency": {
-            "p25": sum(itl_p25_values) / len(itl_p25_values),
-            "p50": sum(itl_p50_values) / len(itl_p50_values),
-            "p75": sum(itl_p75_values) / len(itl_p75_values),
-        },
-    }
-
-    return aggregated
-
-
 def run_benchmark(
     model,
-    tokenizer,  # Add tokenizer parameter
+    tokenizer,
     prefix_ratio,
     isl,
     osl,
@@ -209,25 +115,20 @@ def run_benchmark(
     seed,
     num_prefix_prompts,
     output_dir,
-    urls,
+    url,
     use_expected_osl=False,
 ) -> Optional[Dict]:
     """Run aiperf benchmark for a specific prefix ratio"""
     logger.info(
-        f"Running benchmark with prefix_ratio={prefix_ratio}, seed={seed}, URLs={urls}"
+        f"Running benchmark with prefix_ratio={prefix_ratio}, seed={seed}, url={url}"
     )
 
-    # If single URL, maintain existing behavior
-    if isinstance(urls, str):
-        urls = [urls]
-
-    if len(urls) == 1:
     artifact_dir = f"{output_dir}/prefix_ratio_{prefix_ratio}_seed_{seed}"
     os.makedirs(artifact_dir, exist_ok=True)
 
-        return run_benchmark_single_url(
+    aiperf_cmd = get_aiperf_cmd(
         model,
-            tokenizer,  # Pass tokenizer parameter
+        tokenizer,
         prefix_ratio,
         isl,
         osl,
@@ -236,95 +137,28 @@ def run_benchmark(
         seed,
         num_prefix_prompts,
         artifact_dir,
-            urls[0],
-            use_expected_osl,
-        )
-
-    # Multiple URLs: split requests and concurrency
-    num_urls = len(urls)
-    base_requests_per_url = requests // num_urls
-    remainder_requests = requests % num_urls
-    base_concurrency_per_url = max(1, concurrency // num_urls)
-
-    # Launch parallel processes
-    processes = []
-    artifact_dirs = []
-
-    for i, url in enumerate(urls):
-        # Distribute remainder requests to first few URLs
-        url_requests = base_requests_per_url + (1 if i < remainder_requests else 0)
-
-        artifact_dir = f"{output_dir}/prefix_ratio_{prefix_ratio}_seed_{seed}_url_{i}"
-        os.makedirs(artifact_dir, exist_ok=True)
-        artifact_dirs.append(artifact_dir)
-
-        aiperf_cmd = get_aiperf_cmd(
-            model,
-            tokenizer,  # Pass tokenizer parameter
-            prefix_ratio,
-            isl,
-            osl,
-            url_requests,
-            base_concurrency_per_url,
-            seed,
-            num_prefix_prompts,
-            artifact_dir,
         url,
         use_expected_osl,
     )
 
-        logger.info(f"Launching process for URL {url}: {' '.join(aiperf_cmd)}")
-
-        # Run process without capturing output - let it stream to terminal
-        process = subprocess.Popen(aiperf_cmd)
-        processes.append((process, url, artifact_dir))
-
-    # Wait for all processes to complete and collect results
-    results: List[Optional[Dict]] = []
-    for process, url, artifact_dir in processes:
-        return_code = process.wait()
-
-        if return_code == 0:
-            logger.info(f"AIPerf completed successfully for URL {url}")
+    logger.info(f"Command: {' '.join(aiperf_cmd)}")
 
     try:
-                aiperf_result = get_aiperf_result(artifact_dir)
-                results.append(aiperf_result)
-            except Exception as e:
-                logger.error(f"Failed to get results for URL {url}: {e}")
-                results.append(None)
-        else:
-            logger.error(f"AIPerf failed for URL {url} with error code: {return_code}")
-            results.append(None)
-
-    # Aggregate results
-    return aggregate_results(results)
+        subprocess.run(aiperf_cmd, check=True)
+        logger.info("AIPerf profiling completed successfully")
+        return get_aiperf_result(artifact_dir)
+    except subprocess.CalledProcessError as e:
+        logger.error(f"AIPerf failed with error code: {e.returncode}")
+        return None
 
 
 def main():
     parser = argparse.ArgumentParser(
         description="Benchmark prefix ratios and plot results"
     )
-    parser.add_argument(
-        "--model",
-        type=str,
-        default="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
-        help="Model name",
-    )
-    parser.add_argument(
-        "--tokenizer",
-        type=str,
-        default="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
-        help="Tokenizer name (defaults to model)",
-    )
-    parser.add_argument(
-        "--url",
-        type=str,
-        nargs="+",  # Accept multiple URLs
-        default=["http://localhost:8000"],
-        # default=["http://localhost:8090", "http://localhost:8090"],
-        help="Server URL(s). Can specify multiple URLs for parallel benchmarking",
-    )
+
+    add_common_args(parser)
+
     parser.add_argument(
         "--output-dir",
         type=str,
@@ -336,7 +170,6 @@ def main():
     parser.add_argument("--osl", type=int, default=200, help="Output sequence length")
     parser.add_argument("--requests", type=int, default=200, help="Number of requests")
     parser.add_argument("--concurrency", type=int, default=20, help="Concurrency level")
-    parser.add_argument("--seed", type=int, default=0, help="Initial random seed")
     parser.add_argument(
         "--prefix-ratios",
         type=float,
@@ -344,13 +177,9 @@ def main():
         default=[0.1, 0.3, 0.5, 0.7, 0.9],
         help="List of prefix ratios to test",
     )
-    parser.add_argument(
-        "--use-expected-osl",
-        action="store_true",
-        help="Pass expected_output_tokens to nvext for router tracking",
-    )
 
     args = parser.parse_args()
+    resolve_tokenizer(args)
 
     # Create output directory
     os.makedirs(args.output_dir, exist_ok=True)
@@ -379,7 +208,7 @@ def main():
             current_seed,
             args.num_prefix_prompts,
             args.output_dir,
-            args.url,  # Now passing list of URLs
+            args.url,
             args.use_expected_osl,
         )
 

benchmarks/router/real_data_benchmark.py

@@ -5,23 +5,20 @@
 
 import argparse
 import json
-import logging
 import os
 import subprocess
 
 import numpy as np
+from common import (
+    DEFAULT_MOONCAKE_BLOCK_SIZE,
+    add_common_args,
+    get_common_aiperf_flags,
+    resolve_tokenizer,
+    setup_logger,
+)
 from prefix_data_generator.synthesizer import Synthesizer
 
-# Setup logging
-logger = logging.getLogger(__name__)
-logger.setLevel(logging.INFO)
-console_handler = logging.StreamHandler()
-console_handler.setLevel(logging.INFO)
-formatter = logging.Formatter(
-    "%(asctime)s - %(name)s - %(levelname)s - %(message)s", "%Y-%m-%d %H:%M:%S"
-)
-console_handler.setFormatter(formatter)
-logger.addHandler(console_handler)
+logger = setup_logger(__name__)
 
 
 def get_aiperf_cmd_for_trace(
@@ -30,20 +27,16 @@ def get_aiperf_cmd_for_trace(
     input_dataset,
     artifact_dir,
     seed,
+    block_size,
     url="http://localhost:8888",
 ):
-    return [
+    cmd = [
         "aiperf",
         "profile",
         "--model",
         model,
         "--tokenizer",
         tokenizer,
-        "--endpoint-type",
-        "chat",
-        "--endpoint",
-        "v1/chat/completions",
-        "--streaming",
         "--url",
         url,
         "--input-file",
@@ -51,15 +44,15 @@ def get_aiperf_cmd_for_trace(
         "--custom-dataset-type",
         "mooncake_trace",
         "--fixed-schedule-auto-offset",
+        "--prompt-input-tokens-block-size",
+        str(block_size),
         "--random-seed",
         str(seed),
         "--artifact-dir",
         artifact_dir,
-        "-H",
-        "Authorization: Bearer NOT USED",
-        "-H",
-        "Accept: text/event-stream",
     ]
+    cmd.extend(get_common_aiperf_flags())
+    return cmd
 
 
 def run_benchmark_with_trace(
@@ -69,6 +62,7 @@ def run_benchmark_with_trace(
     artifact_dir,
     url,
     seed,
+    block_size,
 ):
     """Run aiperf benchmark with a trace dataset"""
     aiperf_cmd = get_aiperf_cmd_for_trace(
@@ -77,6 +71,7 @@ def run_benchmark_with_trace(
         trace_dataset,
         artifact_dir,
         seed,
+        block_size,
         url,
     )
 
@@ -100,25 +95,9 @@ def main():
         description="Benchmark with real or synthesized mooncake-style trace data"
     )
 
-    # Model and server configuration
-    parser.add_argument(
-        "--model",
-        type=str,
-        default="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
-        help="Model name",
-    )
-    parser.add_argument(
-        "--tokenizer",
-        type=str,
-        default=None,
-        help="Tokenizer name (defaults to model)",
-    )
-    parser.add_argument(
-        "--url",
-        type=str,
-        default="http://localhost:8000",
-        help="Server URL",
-    )
+    # Common arguments
+    add_common_args(parser)
+
     parser.add_argument(
         "--output-dir",
         type=str,
@@ -190,26 +169,12 @@ def main():
     parser.add_argument(
         "--block-size",
         type=int,
-        default=512,
-        help="Block size for prefilling and decoding (default: 512)",
-    )
-    parser.add_argument(
-        "--seed",
-        type=int,
-        default=0,
-        help="Random seed for reproducibility (default: 0)",
-    )
-    parser.add_argument(
-        "--use-expected-osl",
-        action="store_true",
-        help="Pass expected_output_tokens to nvext for router tracking",
+        default=DEFAULT_MOONCAKE_BLOCK_SIZE,
+        help=f"Block size for prefilling and decoding (default: {DEFAULT_MOONCAKE_BLOCK_SIZE})",
     )
 
     args = parser.parse_args()
-
-    # Use tokenizer from model if not specified
-    if args.tokenizer is None:
-        args.tokenizer = args.model
+    resolve_tokenizer(args)
 
     # Create output directory
     os.makedirs(args.output_dir, exist_ok=True)
@@ -350,6 +315,7 @@ def main():
         artifact_dir,
         args.url,
         args.seed,
+        args.block_size,
     )
 
     logger.info(f"Results saved to: {artifact_dir}")