phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-11 13:44:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

chore(perf): run guidellm benchmarks (#3421)
Some checks failed
Integration Tests (Replay) / Integration Tests (, , , client=, ) (push) Failing after 3s
Details
Unit Tests / unit-tests (3.13) (push) Failing after 3s
Details
Update ReadTheDocs / update-readthedocs (push) Failing after 3s
Details
Test Llama Stack Build / build (push) Failing after 3s
Details
Integration Auth Tests / test-matrix (oauth2_token) (push) Failing after 1s
Details
SqlStore Integration Tests / test-postgres (3.12) (push) Failing after 1s
Details
SqlStore Integration Tests / test-postgres (3.13) (push) Failing after 0s
Details
Test External Providers Installed via Module / test-external-providers-from-module (venv) (push) Has been skipped
Details
Test Llama Stack Build / generate-matrix (push) Successful in 3s
Details
Python Package Build Test / build (3.12) (push) Failing after 1s
Details
Python Package Build Test / build (3.13) (push) Failing after 2s
Details
Test Llama Stack Build / build-custom-container-distribution (push) Failing after 3s
Details
Test Llama Stack Build / build-single-provider (push) Failing after 3s
Details
Vector IO Integration Tests / test-matrix (push) Failing after 5s
Details
Test Llama Stack Build / build-ubi9-container-distribution (push) Failing after 4s
Details
API Conformance Tests / check-schema-compatibility (push) Successful in 8s
Details
Test External API and Providers / test-external (venv) (push) Failing after 3s
Details
Unit Tests / unit-tests (3.12) (push) Failing after 4s
Details
UI Tests / ui-tests (22) (push) Successful in 40s
Details
Pre-commit / pre-commit (push) Successful in 1m9s
Details

Browse source 
# What does this PR do?
- Mostly AI-generated scripts to run guidellm
(https://github.com/vllm-project/guidellm) benchmarks on k8s setup
- Stack is using image built from main on 9/11


## Test Plan
See updated README.md
This commit is contained in:
ehhuang 2025-09-24 10:18:33 -07:00 • committed by GitHub
parent 2f58d87c22
commit 48a551ecbc
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
14 changed files with 1436 additions and 526 deletions
Download patch file Download diff file Expand all files Collapse all files

171
benchmarking/k8s-benchmark/results/guidellm-benchmark-stack-s1-sw1-v1-20250922-103408.txt Normal file
View file

@ -0,0 +1,171 @@
Collecting uv
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (11 kB)
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (20.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 20.9/20.9 MB 144.3 MB/s eta 0:00:00
Installing collected packages: uv
Successfully installed uv-0.8.19
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[notice] A new release of pip is available: 24.0 -> 25.2
[notice] To update, run: pip install --upgrade pip
Using Python 3.11.13 environment at: /usr/local
Resolved 61 packages in 551ms
Downloading pillow (6.3MiB)
Downloading hf-xet (3.0MiB)
Downloading tokenizers (3.1MiB)
Downloading pygments (1.2MiB)
Downloading pandas (11.8MiB)
Downloading aiohttp (1.7MiB)
Downloading pydantic-core (1.9MiB)
Downloading numpy (16.2MiB)
Downloading transformers (11.1MiB)
Downloading pyarrow (40.8MiB)
Downloading pydantic-core
Downloading aiohttp
Downloading tokenizers
Downloading hf-xet
Downloading pygments
Downloading pillow
Downloading numpy
Downloading pandas
Downloading transformers
Downloading pyarrow
Prepared 61 packages in 1.23s
Installed 61 packages in 114ms
+ aiohappyeyeballs==2.6.1
+ aiohttp==3.12.15
+ aiosignal==1.4.0
+ annotated-types==0.7.0
+ anyio==4.10.0
+ attrs==25.3.0
+ certifi==2025.8.3
+ charset-normalizer==3.4.3
+ click==8.1.8
+ datasets==4.1.1
+ dill==0.4.0
+ filelock==3.19.1
+ frozenlist==1.7.0
+ fsspec==2025.9.0
+ ftfy==6.3.1
+ guidellm==0.3.0
+ h11==0.16.0
+ h2==4.3.0
+ hf-xet==1.1.10
+ hpack==4.1.0
+ httpcore==1.0.9
+ httpx==0.28.1
+ huggingface-hub==0.35.0
+ hyperframe==6.1.0
+ idna==3.10
+ loguru==0.7.3
+ markdown-it-py==4.0.0
+ mdurl==0.1.2
+ multidict==6.6.4
+ multiprocess==0.70.16
+ numpy==2.3.3
+ packaging==25.0
+ pandas==2.3.2
+ pillow==11.3.0
+ propcache==0.3.2
+ protobuf==6.32.1
+ pyarrow==21.0.0
+ pydantic==2.11.9
+ pydantic-core==2.33.2
+ pydantic-settings==2.10.1
+ pygments==2.19.2
+ python-dateutil==2.9.0.post0
+ python-dotenv==1.1.1
+ pytz==2025.2
+ pyyaml==6.0.2
+ regex==2025.9.18
+ requests==2.32.5
+ rich==14.1.0
+ safetensors==0.6.2
+ six==1.17.0
+ sniffio==1.3.1
+ tokenizers==0.22.1
+ tqdm==4.67.1
+ transformers==4.56.2
+ typing-extensions==4.15.0
+ typing-inspection==0.4.1
+ tzdata==2025.2
+ urllib3==2.5.0
+ wcwidth==0.2.14
+ xxhash==3.5.0
+ yarl==1.20.1
Using Python 3.11.13 environment at: /usr/local
Audited 1 package in 3ms
Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
Creating backend...
Backend openai_http connected to http://llama-stack-benchmark-service:8323/v1/openai for model meta-llama/Llama-3.2-3B-Instruct.
Creating request loader...
Created loader with 1000 unique requests from prompt_tokens=512,output_tokens=256.
╭─ Benchmarks ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ [17:34:30] ⠋ 100% concurrent@1 (complete) Req: 0.3 req/s, 3.32s Lat, 1.0 Conc, 18 Comp, 1 Inc, 0 Err │
│ Tok: 74.0 gen/s, 238.6 tot/s, 40.2ms TTFT, 13.4ms ITL, 546 Prompt, 246 Gen │
│ [17:35:35] ⠋ 100% concurrent@2 (complete) Req: 0.6 req/s, 3.46s Lat, 2.0 Conc, 34 Comp, 2 Inc, 0 Err │
│ Tok: 139.6 gen/s, 454.0 tot/s, 48.0ms TTFT, 14.1ms ITL, 546 Prompt, 243 Gen │
│ [17:36:40] ⠋ 100% concurrent@4 (complete) Req: 1.1 req/s, 3.44s Lat, 3.9 Conc, 68 Comp, 4 Inc, 0 Err │
│ Tok: 273.2 gen/s, 900.4 tot/s, 50.7ms TTFT, 14.3ms ITL, 546 Prompt, 238 Gen │
│ [17:37:45] ⠋ 100% concurrent@8 (complete) Req: 2.2 req/s, 3.55s Lat, 7.7 Conc, 129 Comp, 8 Inc, 0 Err │
│ Tok: 519.1 gen/s, 1699.8 tot/s, 66.0ms TTFT, 14.6ms ITL, 547 Prompt, 240 Gen │
│ [17:38:50] ⠋ 100% concurrent@16 (complete) Req: 4.1 req/s, 3.76s Lat, 15.5 Conc, 247 Comp, 16 Inc, 0 Err │
│ Tok: 1005.5 gen/s, 3256.7 tot/s, 101.0ms TTFT, 15.0ms ITL, 547 Prompt, 244 Gen │
│ [17:39:56] ⠋ 100% concurrent@32 (complete) Req: 8.1 req/s, 3.84s Lat, 30.9 Conc, 483 Comp, 32 Inc, 0 Err │
│ Tok: 1926.3 gen/s, 6327.2 tot/s, 295.7ms TTFT, 14.8ms ITL, 547 Prompt, 239 Gen │
│ [17:41:03] ⠋ 100% concurrent@64 (complete) Req: 9.9 req/s, 6.05s Lat, 59.7 Conc, 576 Comp, 58 Inc, 0 Err │
│ Tok: 2381.0 gen/s, 7774.5 tot/s, 1196.2ms TTFT, 20.2ms ITL, 547 Prompt, 241 Gen │
│ [17:42:10] ⠋ 100% concurrent@128 (complete) Req: 9.2 req/s, 11.59s Lat, 107.2 Conc, 514 Comp, 117 Inc, 0 Err │
│ Tok: 2233.4 gen/s, 7286.3 tot/s, 2403.9ms TTFT, 38.2ms ITL, 547 Prompt, 242 Gen │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
Generating... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ (8/8) [ 0:08:41 < 0:00:00 ]
Benchmarks Metadata:
Run id:511a14fd-ba11-4ffa-92ef-7cc23db4dd38
Duration:528.5 seconds
Profile:type=concurrent, strategies=['concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent'], streams=[1, 2, 4, 8, 16, 32, 64, 128]
Args:max_number=None, max_duration=60.0, warmup_number=None, warmup_duration=3.0, cooldown_number=None, cooldown_duration=None
Worker:type_='generative_requests_worker' backend_type='openai_http' backend_target='http://llama-stack-benchmark-service:8323/v1/openai' backend_model='meta-llama/Llama-3.2-3B-Instruct'
backend_info={'max_output_tokens': 16384, 'timeout': 300, 'http2': True, 'follow_redirects': True, 'headers': {}, 'text_completions_path': '/v1/completions', 'chat_completions_path':
'/v1/chat/completions'}
Request Loader:type_='generative_request_loader' data='prompt_tokens=512,output_tokens=256' data_args=None processor='meta-llama/Llama-3.2-3B-Instruct' processor_args=None
Extras:None
Benchmarks Info:
===================================================================================================================================================
Metadata |||| Requests Made ||| Prompt Tok/Req ||| Output Tok/Req ||| Prompt Tok Total||| Output Tok Total||
Benchmark| Start Time| End Time| Duration (s)| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err
--------------|-----------|---------|-------------|------|-----|-----|------|------|----|------|------|----|-------|------|----|-------|------|----
concurrent@1| 17:34:35| 17:35:35| 60.0| 18| 1| 0| 546.4| 512.0| 0.0| 246.0| 14.0| 0.0| 9835| 512| 0| 4428| 14| 0
concurrent@2| 17:35:40| 17:36:40| 60.0| 34| 2| 0| 546.4| 512.0| 0.0| 242.7| 80.0| 0.0| 18577| 1024| 0| 8253| 160| 0
concurrent@4| 17:36:45| 17:37:45| 60.0| 68| 4| 0| 546.4| 512.0| 0.0| 238.1| 103.2| 0.0| 37156| 2048| 0| 16188| 413| 0
concurrent@8| 17:37:50| 17:38:50| 60.0| 129| 8| 0| 546.7| 512.0| 0.0| 240.3| 180.0| 0.0| 70518| 4096| 0| 31001| 1440| 0
concurrent@16| 17:38:55| 17:39:55| 60.0| 247| 16| 0| 546.6| 512.0| 0.0| 244.1| 142.6| 0.0| 135002| 8192| 0| 60300| 2281| 0
concurrent@32| 17:40:01| 17:41:01| 60.0| 483| 32| 0| 546.5| 512.0| 0.0| 239.2| 123.2| 0.0| 263972| 16384| 0| 115540| 3944| 0
concurrent@64| 17:41:08| 17:42:08| 60.0| 576| 58| 0| 546.6| 512.0| 0.0| 241.3| 13.9| 0.0| 314817| 29696| 0| 138976| 807| 0
concurrent@128| 17:42:15| 17:43:15| 60.0| 514| 117| 0| 546.5| 512.0| 0.0| 241.6| 143.9| 0.0| 280911| 59904| 0| 124160| 16832| 0
===================================================================================================================================================
Benchmarks Stats:
=======================================================================================================================================================
Metadata | Request Stats || Out Tok/sec| Tot Tok/sec| Req Latency (sec) ||| TTFT (ms) ||| ITL (ms) ||| TPOT (ms) ||
Benchmark| Per Second| Concurrency| mean| mean| mean| median| p99| mean| median| p99| mean| median| p99| mean| median| p99
--------------|-----------|------------|------------|------------|------|-------|------|-------|-------|-------|-----|-------|-----|-----|-------|-----
concurrent@1| 0.30| 1.00| 74.0| 238.6| 3.32| 3.43| 3.61| 40.2| 39.3| 51.2| 13.4| 13.3| 14.0| 13.3| 13.2| 13.9
concurrent@2| 0.58| 1.99| 139.6| 454.0| 3.46| 3.64| 3.74| 48.0| 45.8| 72.0| 14.1| 14.1| 14.5| 14.0| 14.0| 14.4
concurrent@4| 1.15| 3.95| 273.2| 900.4| 3.44| 3.69| 3.74| 50.7| 47.2| 118.6| 14.3| 14.3| 14.4| 14.2| 14.2| 14.4
concurrent@8| 2.16| 7.67| 519.1| 1699.8| 3.55| 3.76| 3.87| 66.0| 48.8| 208.2| 14.6| 14.5| 14.8| 14.5| 14.5| 14.8
concurrent@16| 4.12| 15.48| 1005.5| 3256.7| 3.76| 3.90| 4.18| 101.0| 65.6| 396.7| 15.0| 15.0| 15.9| 15.0| 15.0| 15.9
concurrent@32| 8.05| 30.89| 1926.3| 6327.2| 3.84| 4.04| 4.39| 295.7| 265.6| 720.4| 14.8| 14.9| 15.5| 14.8| 14.8| 15.3
concurrent@64| 9.87| 59.74| 2381.0| 7774.5| 6.05| 6.18| 9.94| 1196.2| 1122.5| 4295.3| 20.2| 20.0| 25.8| 20.1| 19.9| 25.8
concurrent@128| 9.25| 107.16| 2233.4| 7286.3| 11.59| 12.04| 14.46| 2403.9| 2322.3| 4001.5| 38.2| 38.5| 53.0| 38.0| 38.3| 52.7
=======================================================================================================================================================
Saving benchmarks report...
Benchmarks report saved to /benchmarks.json
Benchmarking complete.

171
benchmarking/k8s-benchmark/results/guidellm-benchmark-stack-s1-sw2-v1-20250922-104457.txt Normal file
View file

@ -0,0 +1,171 @@
Collecting uv
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (11 kB)
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (20.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 20.9/20.9 MB 149.3 MB/s eta 0:00:00
Installing collected packages: uv
Successfully installed uv-0.8.19
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[notice] A new release of pip is available: 24.0 -> 25.2
[notice] To update, run: pip install --upgrade pip
Using Python 3.11.13 environment at: /usr/local
Resolved 61 packages in 494ms
Downloading pandas (11.8MiB)
Downloading tokenizers (3.1MiB)
Downloading pygments (1.2MiB)
Downloading aiohttp (1.7MiB)
Downloading transformers (11.1MiB)
Downloading numpy (16.2MiB)
Downloading pillow (6.3MiB)
Downloading pydantic-core (1.9MiB)
Downloading hf-xet (3.0MiB)
Downloading pyarrow (40.8MiB)
Downloading pydantic-core
Downloading aiohttp
Downloading tokenizers
Downloading hf-xet
Downloading pillow
Downloading pygments
Downloading numpy
Downloading pandas
Downloading pyarrow
Downloading transformers
Prepared 61 packages in 1.24s
Installed 61 packages in 126ms
+ aiohappyeyeballs==2.6.1
+ aiohttp==3.12.15
+ aiosignal==1.4.0
+ annotated-types==0.7.0
+ anyio==4.10.0
+ attrs==25.3.0
+ certifi==2025.8.3
+ charset-normalizer==3.4.3
+ click==8.1.8
+ datasets==4.1.1
+ dill==0.4.0
+ filelock==3.19.1
+ frozenlist==1.7.0
+ fsspec==2025.9.0
+ ftfy==6.3.1
+ guidellm==0.3.0
+ h11==0.16.0
+ h2==4.3.0
+ hf-xet==1.1.10
+ hpack==4.1.0
+ httpcore==1.0.9
+ httpx==0.28.1
+ huggingface-hub==0.35.0
+ hyperframe==6.1.0
+ idna==3.10
+ loguru==0.7.3
+ markdown-it-py==4.0.0
+ mdurl==0.1.2
+ multidict==6.6.4
+ multiprocess==0.70.16
+ numpy==2.3.3
+ packaging==25.0
+ pandas==2.3.2
+ pillow==11.3.0
+ propcache==0.3.2
+ protobuf==6.32.1
+ pyarrow==21.0.0
+ pydantic==2.11.9
+ pydantic-core==2.33.2
+ pydantic-settings==2.10.1
+ pygments==2.19.2
+ python-dateutil==2.9.0.post0
+ python-dotenv==1.1.1
+ pytz==2025.2
+ pyyaml==6.0.2
+ regex==2025.9.18
+ requests==2.32.5
+ rich==14.1.0
+ safetensors==0.6.2
+ six==1.17.0
+ sniffio==1.3.1
+ tokenizers==0.22.1
+ tqdm==4.67.1
+ transformers==4.56.2
+ typing-extensions==4.15.0
+ typing-inspection==0.4.1
+ tzdata==2025.2
+ urllib3==2.5.0
+ wcwidth==0.2.14
+ xxhash==3.5.0
+ yarl==1.20.1
Using Python 3.11.13 environment at: /usr/local
Audited 1 package in 3ms
Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
Creating backend...
Backend openai_http connected to http://llama-stack-benchmark-service:8323/v1/openai for model meta-llama/Llama-3.2-3B-Instruct.
Creating request loader...
Created loader with 1000 unique requests from prompt_tokens=512,output_tokens=256.
╭─ Benchmarks ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ [17:45:18] ⠋ 100% concurrent@1 (complete) Req: 0.3 req/s, 3.42s Lat, 1.0 Conc, 17 Comp, 1 Inc, 0 Err │
│ Tok: 73.9 gen/s, 233.7 tot/s, 50.2ms TTFT, 13.4ms ITL, 547 Prompt, 253 Gen │
│ [17:46:23] ⠋ 100% concurrent@2 (complete) Req: 0.6 req/s, 3.42s Lat, 2.0 Conc, 34 Comp, 2 Inc, 0 Err │
│ Tok: 134.7 gen/s, 447.4 tot/s, 50.8ms TTFT, 14.3ms ITL, 546 Prompt, 235 Gen │
│ [17:47:28] ⠋ 100% concurrent@4 (complete) Req: 1.1 req/s, 3.55s Lat, 3.9 Conc, 66 Comp, 4 Inc, 0 Err │
│ Tok: 268.7 gen/s, 873.1 tot/s, 54.9ms TTFT, 14.4ms ITL, 547 Prompt, 243 Gen │
│ [17:48:33] ⠋ 100% concurrent@8 (complete) Req: 2.2 req/s, 3.56s Lat, 7.8 Conc, 130 Comp, 8 Inc, 0 Err │
│ Tok: 526.1 gen/s, 1728.4 tot/s, 60.6ms TTFT, 14.7ms ITL, 547 Prompt, 239 Gen │
│ [17:49:38] ⠋ 100% concurrent@16 (complete) Req: 4.1 req/s, 3.79s Lat, 15.7 Conc, 246 Comp, 16 Inc, 0 Err │
│ Tok: 1006.9 gen/s, 3268.6 tot/s, 74.8ms TTFT, 15.3ms ITL, 547 Prompt, 243 Gen │
│ [17:50:44] ⠋ 100% concurrent@32 (complete) Req: 7.8 req/s, 3.95s Lat, 30.9 Conc, 467 Comp, 32 Inc, 0 Err │
│ Tok: 1912.0 gen/s, 6191.6 tot/s, 119.1ms TTFT, 15.7ms ITL, 547 Prompt, 244 Gen │
│ [17:51:50] ⠋ 100% concurrent@64 (complete) Req: 13.0 req/s, 4.75s Lat, 61.8 Conc, 776 Comp, 64 Inc, 0 Err │
│ Tok: 3154.3 gen/s, 10273.3 tot/s, 339.1ms TTFT, 18.3ms ITL, 547 Prompt, 242 Gen │
│ [17:52:58] ⠋ 100% concurrent@128 (complete) Req: 15.1 req/s, 7.82s Lat, 117.7 Conc, 898 Comp, 127 Inc, 0 Err │
│ Tok: 3617.4 gen/s, 11843.9 tot/s, 1393.8ms TTFT, 26.8ms ITL, 547 Prompt, 240 Gen │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
Generating... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ (8/8) [ 0:08:41 < 0:00:00 ]
Benchmarks Metadata:
Run id:f73d408e-256a-4c32-aa40-05e8d7098b66
Duration:529.2 seconds
Profile:type=concurrent, strategies=['concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent'], streams=[1, 2, 4, 8, 16, 32, 64, 128]
Args:max_number=None, max_duration=60.0, warmup_number=None, warmup_duration=3.0, cooldown_number=None, cooldown_duration=None
Worker:type_='generative_requests_worker' backend_type='openai_http' backend_target='http://llama-stack-benchmark-service:8323/v1/openai' backend_model='meta-llama/Llama-3.2-3B-Instruct'
backend_info={'max_output_tokens': 16384, 'timeout': 300, 'http2': True, 'follow_redirects': True, 'headers': {}, 'text_completions_path': '/v1/completions', 'chat_completions_path':
'/v1/chat/completions'}
Request Loader:type_='generative_request_loader' data='prompt_tokens=512,output_tokens=256' data_args=None processor='meta-llama/Llama-3.2-3B-Instruct' processor_args=None
Extras:None
Benchmarks Info:
=====================================================================================================================================================
Metadata |||| Requests Made ||| Prompt Tok/Req ||| Output Tok/Req ||| Prompt Tok Total||| Output Tok Total ||
Benchmark| Start Time| End Time| Duration (s)| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err
--------------|-----------|---------|-------------|------|-----|-----|------|------|----|------|------|----|-------|------|----|--------|------|-----
concurrent@1| 17:45:23| 17:46:23| 60.0| 17| 1| 0| 546.6| 512.0| 0.0| 252.8| 136.0| 0.0| 9292| 512| 0| 4298| 136| 0
concurrent@2| 17:46:28| 17:47:28| 60.0| 34| 2| 0| 546.4| 512.0| 0.0| 235.4| 130.0| 0.0| 18577| 1024| 0| 8003| 260| 0
concurrent@4| 17:47:33| 17:48:33| 60.0| 66| 4| 0| 546.5| 512.0| 0.0| 243.0| 97.5| 0.0| 36072| 2048| 0| 16035| 390| 0
concurrent@8| 17:48:38| 17:49:38| 60.0| 130| 8| 0| 546.6| 512.0| 0.0| 239.2| 146.0| 0.0| 71052| 4096| 0| 31090| 1168| 0
concurrent@16| 17:49:43| 17:50:43| 60.0| 246| 16| 0| 546.6| 512.0| 0.0| 243.3| 112.3| 0.0| 134456| 8192| 0| 59862| 1797| 0
concurrent@32| 17:50:49| 17:51:49| 60.0| 467| 32| 0| 546.6| 512.0| 0.0| 244.2| 147.3| 0.0| 255242| 16384| 0| 114038| 4714| 0
concurrent@64| 17:51:55| 17:52:55| 60.0| 776| 64| 0| 546.5| 512.0| 0.0| 242.2| 106.1| 0.0| 424115| 32768| 0| 187916| 6788| 0
concurrent@128| 17:53:03| 17:54:03| 60.0| 898| 127| 0| 546.5| 512.0| 0.0| 240.3| 69.8| 0.0| 490789| 65024| 0| 215810| 8864| 0
=====================================================================================================================================================
Benchmarks Stats:
======================================================================================================================================================
Metadata | Request Stats || Out Tok/sec| Tot Tok/sec| Req Latency (sec)||| TTFT (ms) ||| ITL (ms) ||| TPOT (ms) ||
Benchmark| Per Second| Concurrency| mean| mean| mean| median| p99| mean| median| p99| mean| median| p99| mean| median| p99
--------------|-----------|------------|------------|------------|-----|-------|------|-------|-------|-------|-----|-------|-----|-----|-------|-----
concurrent@1| 0.29| 1.00| 73.9| 233.7| 3.42| 3.45| 3.50| 50.2| 50.9| 62.5| 13.4| 13.4| 13.5| 13.3| 13.3| 13.5
concurrent@2| 0.57| 1.96| 134.7| 447.4| 3.42| 3.67| 4.12| 50.8| 49.2| 79.8| 14.3| 14.2| 15.9| 14.3| 14.2| 15.9
concurrent@4| 1.11| 3.92| 268.7| 873.1| 3.55| 3.72| 3.80| 54.9| 51.7| 101.3| 14.4| 14.4| 14.5| 14.4| 14.4| 14.5
concurrent@8| 2.20| 7.82| 526.1| 1728.4| 3.56| 3.78| 3.93| 60.6| 49.8| 189.5| 14.7| 14.7| 14.8| 14.6| 14.6| 14.8
concurrent@16| 4.14| 15.66| 1006.9| 3268.6| 3.79| 3.94| 4.25| 74.8| 54.3| 328.4| 15.3| 15.3| 16.1| 15.2| 15.2| 16.0
concurrent@32| 7.83| 30.91| 1912.0| 6191.6| 3.95| 4.07| 4.53| 119.1| 80.5| 674.0| 15.7| 15.6| 17.4| 15.7| 15.6| 17.3
concurrent@64| 13.03| 61.85| 3154.3| 10273.3| 4.75| 4.93| 5.43| 339.1| 321.1| 1146.6| 18.3| 18.4| 19.3| 18.2| 18.3| 19.2
concurrent@128| 15.05| 117.71| 3617.4| 11843.9| 7.82| 8.58| 13.35| 1393.8| 1453.0| 5232.2| 26.8| 26.7| 36.0| 26.7| 26.6| 35.9
======================================================================================================================================================
Saving benchmarks report...
Benchmarks report saved to /benchmarks.json
Benchmarking complete.

171
benchmarking/k8s-benchmark/results/guidellm-benchmark-stack-s1-sw4-v1-20250922-105539.txt Normal file
View file

@ -0,0 +1,171 @@
Collecting uv
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (11 kB)
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (20.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 20.9/20.9 MB 156.8 MB/s eta 0:00:00
Installing collected packages: uv
Successfully installed uv-0.8.19
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[notice] A new release of pip is available: 24.0 -> 25.2
[notice] To update, run: pip install --upgrade pip
Using Python 3.11.13 environment at: /usr/local
Resolved 61 packages in 480ms
Downloading pillow (6.3MiB)
Downloading pydantic-core (1.9MiB)
Downloading pyarrow (40.8MiB)
Downloading aiohttp (1.7MiB)
Downloading numpy (16.2MiB)
Downloading pygments (1.2MiB)
Downloading transformers (11.1MiB)
Downloading pandas (11.8MiB)
Downloading tokenizers (3.1MiB)
Downloading hf-xet (3.0MiB)
Downloading pydantic-core
Downloading aiohttp
Downloading tokenizers
Downloading hf-xet
Downloading pygments
Downloading pillow
Downloading numpy
Downloading pandas
Downloading pyarrow
Downloading transformers
Prepared 61 packages in 1.25s
Installed 61 packages in 126ms
+ aiohappyeyeballs==2.6.1
+ aiohttp==3.12.15
+ aiosignal==1.4.0
+ annotated-types==0.7.0
+ anyio==4.10.0
+ attrs==25.3.0
+ certifi==2025.8.3
+ charset-normalizer==3.4.3
+ click==8.1.8
+ datasets==4.1.1
+ dill==0.4.0
+ filelock==3.19.1
+ frozenlist==1.7.0
+ fsspec==2025.9.0
+ ftfy==6.3.1
+ guidellm==0.3.0
+ h11==0.16.0
+ h2==4.3.0
+ hf-xet==1.1.10
+ hpack==4.1.0
+ httpcore==1.0.9
+ httpx==0.28.1
+ huggingface-hub==0.35.0
+ hyperframe==6.1.0
+ idna==3.10
+ loguru==0.7.3
+ markdown-it-py==4.0.0
+ mdurl==0.1.2
+ multidict==6.6.4
+ multiprocess==0.70.16
+ numpy==2.3.3
+ packaging==25.0
+ pandas==2.3.2
+ pillow==11.3.0
+ propcache==0.3.2
+ protobuf==6.32.1
+ pyarrow==21.0.0
+ pydantic==2.11.9
+ pydantic-core==2.33.2
+ pydantic-settings==2.10.1
+ pygments==2.19.2
+ python-dateutil==2.9.0.post0
+ python-dotenv==1.1.1
+ pytz==2025.2
+ pyyaml==6.0.2
+ regex==2025.9.18
+ requests==2.32.5
+ rich==14.1.0
+ safetensors==0.6.2
+ six==1.17.0
+ sniffio==1.3.1
+ tokenizers==0.22.1
+ tqdm==4.67.1
+ transformers==4.56.2
+ typing-extensions==4.15.0
+ typing-inspection==0.4.1
+ tzdata==2025.2
+ urllib3==2.5.0
+ wcwidth==0.2.14
+ xxhash==3.5.0
+ yarl==1.20.1
Using Python 3.11.13 environment at: /usr/local
Audited 1 package in 4ms
Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
Creating backend...
Backend openai_http connected to http://llama-stack-benchmark-service:8323/v1/openai for model meta-llama/Llama-3.2-3B-Instruct.
Creating request loader...
Created loader with 1000 unique requests from prompt_tokens=512,output_tokens=256.
╭─ Benchmarks ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ [17:55:59] ⠋ 100% concurrent@1 (complete) Req: 0.3 req/s, 3.33s Lat, 1.0 Conc, 18 Comp, 1 Inc, 0 Err │
│ Tok: 74.0 gen/s, 238.0 tot/s, 49.6ms TTFT, 13.4ms ITL, 546 Prompt, 246 Gen │
│ [17:57:04] ⠋ 100% concurrent@2 (complete) Req: 0.6 req/s, 3.32s Lat, 1.9 Conc, 35 Comp, 2 Inc, 0 Err │
│ Tok: 137.1 gen/s, 457.5 tot/s, 50.6ms TTFT, 14.0ms ITL, 546 Prompt, 234 Gen │
│ [17:58:09] ⠋ 100% concurrent@4 (complete) Req: 1.2 req/s, 3.42s Lat, 4.0 Conc, 69 Comp, 4 Inc, 0 Err │
│ Tok: 276.7 gen/s, 907.2 tot/s, 52.7ms TTFT, 14.1ms ITL, 547 Prompt, 240 Gen │
│ [17:59:14] ⠋ 100% concurrent@8 (complete) Req: 2.3 req/s, 3.47s Lat, 7.8 Conc, 134 Comp, 8 Inc, 0 Err │
│ Tok: 541.4 gen/s, 1775.4 tot/s, 57.3ms TTFT, 14.3ms ITL, 547 Prompt, 240 Gen │
│ [18:00:19] ⠋ 100% concurrent@16 (complete) Req: 4.3 req/s, 3.60s Lat, 15.6 Conc, 259 Comp, 16 Inc, 0 Err │
│ Tok: 1034.8 gen/s, 3401.7 tot/s, 72.3ms TTFT, 14.8ms ITL, 547 Prompt, 239 Gen │
│ [18:01:25] ⠋ 100% concurrent@32 (complete) Req: 8.4 req/s, 3.69s Lat, 31.1 Conc, 505 Comp, 32 Inc, 0 Err │
│ Tok: 2029.7 gen/s, 6641.5 tot/s, 91.6ms TTFT, 15.0ms ITL, 547 Prompt, 241 Gen │
│ [18:02:31] ⠋ 100% concurrent@64 (complete) Req: 13.6 req/s, 4.50s Lat, 61.4 Conc, 818 Comp, 64 Inc, 0 Err │
│ Tok: 3333.9 gen/s, 10787.0 tot/s, 171.3ms TTFT, 17.8ms ITL, 547 Prompt, 244 Gen │
│ [18:03:40] ⠋ 100% concurrent@128 (complete) Req: 16.1 req/s, 7.43s Lat, 119.5 Conc, 964 Comp, 122 Inc, 0 Err │
│ Tok: 3897.0 gen/s, 12679.4 tot/s, 446.4ms TTFT, 28.9ms ITL, 547 Prompt, 243 Gen │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
Generating... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ (8/8) [ 0:08:41 < 0:00:00 ]
Benchmarks Metadata:
Run id:5393e64f-d9f8-4548-95d8-da320bba1c24
Duration:530.1 seconds
Profile:type=concurrent, strategies=['concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent'], streams=[1, 2, 4, 8, 16, 32, 64, 128]
Args:max_number=None, max_duration=60.0, warmup_number=None, warmup_duration=3.0, cooldown_number=None, cooldown_duration=None
Worker:type_='generative_requests_worker' backend_type='openai_http' backend_target='http://llama-stack-benchmark-service:8323/v1/openai' backend_model='meta-llama/Llama-3.2-3B-Instruct'
backend_info={'max_output_tokens': 16384, 'timeout': 300, 'http2': True, 'follow_redirects': True, 'headers': {}, 'text_completions_path': '/v1/completions', 'chat_completions_path':
'/v1/chat/completions'}
Request Loader:type_='generative_request_loader' data='prompt_tokens=512,output_tokens=256' data_args=None processor='meta-llama/Llama-3.2-3B-Instruct' processor_args=None
Extras:None
Benchmarks Info:
===================================================================================================================================================
Metadata |||| Requests Made ||| Prompt Tok/Req ||| Output Tok/Req ||| Prompt Tok Total||| Output Tok Total||
Benchmark| Start Time| End Time| Duration (s)| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err
--------------|-----------|---------|-------------|------|-----|-----|------|------|----|------|------|----|-------|------|----|-------|------|----
concurrent@1| 17:56:04| 17:57:04| 60.0| 18| 1| 0| 546.4| 512.0| 0.0| 246.4| 256.0| 0.0| 9836| 512| 0| 4436| 256| 0
concurrent@2| 17:57:09| 17:58:09| 60.0| 35| 2| 0| 546.4| 512.0| 0.0| 233.9| 132.0| 0.0| 19124| 1024| 0| 8188| 264| 0
concurrent@4| 17:58:14| 17:59:14| 60.0| 69| 4| 0| 546.6| 512.0| 0.0| 239.9| 60.5| 0.0| 37715| 2048| 0| 16553| 242| 0
concurrent@8| 17:59:19| 18:00:19| 60.0| 134| 8| 0| 546.6| 512.0| 0.0| 239.8| 126.6| 0.0| 73243| 4096| 0| 32135| 1013| 0
concurrent@16| 18:00:24| 18:01:24| 60.0| 259| 16| 0| 546.6| 512.0| 0.0| 239.0| 115.7| 0.0| 141561| 8192| 0| 61889| 1851| 0
concurrent@32| 18:01:30| 18:02:30| 60.0| 505| 32| 0| 546.5| 512.0| 0.0| 240.5| 113.2| 0.0| 275988| 16384| 0| 121466| 3623| 0
concurrent@64| 18:02:37| 18:03:37| 60.0| 818| 64| 0| 546.6| 512.0| 0.0| 244.5| 132.4| 0.0| 447087| 32768| 0| 199988| 8475| 0
concurrent@128| 18:03:45| 18:04:45| 60.0| 964| 122| 0| 546.5| 512.0| 0.0| 242.5| 133.1| 0.0| 526866| 62464| 0| 233789| 16241| 0
===================================================================================================================================================
Benchmarks Stats:
=======================================================================================================================================================
Metadata | Request Stats || Out Tok/sec| Tot Tok/sec| Req Latency (sec) ||| TTFT (ms) ||| ITL (ms) ||| TPOT (ms) ||
Benchmark| Per Second| Concurrency| mean| mean| mean| median| p99| mean| median| p99| mean| median| p99| mean| median| p99
--------------|-----------|------------|------------|------------|------|--------|------|------|-------|-------|-----|-------|-----|-----|-------|-----
concurrent@1| 0.30| 1.00| 74.0| 238.0| 3.33| 3.44| 3.63| 49.6| 47.2| 66.1| 13.4| 13.3| 14.0| 13.3| 13.3| 14.0
concurrent@2| 0.59| 1.95| 137.1| 457.5| 3.32| 3.61| 3.67| 50.6| 48.6| 80.4| 14.0| 14.0| 14.2| 13.9| 13.9| 14.1
concurrent@4| 1.15| 3.95| 276.7| 907.2| 3.42| 3.61| 3.77| 52.7| 49.7| 106.9| 14.1| 14.0| 14.6| 14.0| 13.9| 14.5
concurrent@8| 2.26| 7.83| 541.4| 1775.4| 3.47| 3.70| 3.79| 57.3| 50.9| 171.3| 14.3| 14.3| 14.4| 14.2| 14.2| 14.4
concurrent@16| 4.33| 15.57| 1034.8| 3401.7| 3.60| 3.81| 4.22| 72.3| 52.0| 292.9| 14.8| 14.7| 16.3| 14.7| 14.7| 16.3
concurrent@32| 8.44| 31.12| 2029.7| 6641.5| 3.69| 3.89| 4.24| 91.6| 62.6| 504.6| 15.0| 15.0| 15.4| 14.9| 14.9| 15.4
concurrent@64| 13.64| 61.40| 3333.9| 10787.0| 4.50| 4.61| 5.67| 171.3| 101.2| 1165.6| 17.8| 17.7| 19.2| 17.7| 17.6| 19.1
concurrent@128| 16.07| 119.45| 3897.0| 12679.4| 7.43| 7.63| 9.74| 446.4| 195.8| 2533.1| 28.9| 28.9| 31.0| 28.8| 28.8| 30.9
=======================================================================================================================================================
Saving benchmarks report...
Benchmarks report saved to /benchmarks.json
Benchmarking complete.

170
benchmarking/k8s-benchmark/results/guidellm-benchmark-vllm-v1-20250922-111127.txt Normal file
View file

@ -0,0 +1,170 @@
Collecting uv
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (11 kB)
Downloading uv-0.8.19-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (20.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 20.9/20.9 MB 126.9 MB/s eta 0:00:00
Installing collected packages: uv
Successfully installed uv-0.8.19
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[notice] A new release of pip is available: 24.0 -> 25.2
[notice] To update, run: pip install --upgrade pip
Using Python 3.11.13 environment at: /usr/local
Resolved 61 packages in 561ms
Downloading hf-xet (3.0MiB)
Downloading pillow (6.3MiB)
Downloading transformers (11.1MiB)
Downloading pyarrow (40.8MiB)
Downloading numpy (16.2MiB)
Downloading pandas (11.8MiB)
Downloading tokenizers (3.1MiB)
Downloading pydantic-core (1.9MiB)
Downloading pygments (1.2MiB)
Downloading aiohttp (1.7MiB)
Downloading pydantic-core
Downloading aiohttp
Downloading tokenizers
Downloading hf-xet
Downloading pygments
Downloading pillow
Downloading numpy
Downloading pandas
Downloading transformers
Downloading pyarrow
Prepared 61 packages in 1.25s
Installed 61 packages in 114ms
+ aiohappyeyeballs==2.6.1
+ aiohttp==3.12.15
+ aiosignal==1.4.0
+ annotated-types==0.7.0
+ anyio==4.10.0
+ attrs==25.3.0
+ certifi==2025.8.3
+ charset-normalizer==3.4.3
+ click==8.1.8
+ datasets==4.1.1
+ dill==0.4.0
+ filelock==3.19.1
+ frozenlist==1.7.0
+ fsspec==2025.9.0
+ ftfy==6.3.1
+ guidellm==0.3.0
+ h11==0.16.0
+ h2==4.3.0
+ hf-xet==1.1.10
+ hpack==4.1.0
+ httpcore==1.0.9
+ httpx==0.28.1
+ huggingface-hub==0.35.0
+ hyperframe==6.1.0
+ idna==3.10
+ loguru==0.7.3
+ markdown-it-py==4.0.0
+ mdurl==0.1.2
+ multidict==6.6.4
+ multiprocess==0.70.16
+ numpy==2.3.3
+ packaging==25.0
+ pandas==2.3.2
+ pillow==11.3.0
+ propcache==0.3.2
+ protobuf==6.32.1
+ pyarrow==21.0.0
+ pydantic==2.11.9
+ pydantic-core==2.33.2
+ pydantic-settings==2.10.1
+ pygments==2.19.2
+ python-dateutil==2.9.0.post0
+ python-dotenv==1.1.1
+ pytz==2025.2
+ pyyaml==6.0.2
+ regex==2025.9.18
+ requests==2.32.5
+ rich==14.1.0
+ safetensors==0.6.2
+ six==1.17.0
+ sniffio==1.3.1
+ tokenizers==0.22.1
+ tqdm==4.67.1
+ transformers==4.56.2
+ typing-extensions==4.15.0
+ typing-inspection==0.4.1
+ tzdata==2025.2
+ urllib3==2.5.0
+ wcwidth==0.2.14
+ xxhash==3.5.0
+ yarl==1.20.1
Using Python 3.11.13 environment at: /usr/local
Audited 1 package in 3ms
Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
Creating backend...
Backend openai_http connected to http://vllm-server:8000 for model meta-llama/Llama-3.2-3B-Instruct.
Creating request loader...
Created loader with 1000 unique requests from prompt_tokens=512,output_tokens=256.
╭─ Benchmarks ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ [18:11:47] ⠋ 100% concurrent@1 (complete) Req: 0.3 req/s, 3.35s Lat, 1.0 Conc, 17 Comp, 1 Inc, 0 Err │
│ Tok: 76.4 gen/s, 239.4 tot/s, 29.6ms TTFT, 13.0ms ITL, 547 Prompt, 256 Gen │
│ [18:12:52] ⠋ 100% concurrent@2 (complete) Req: 0.6 req/s, 3.53s Lat, 2.0 Conc, 32 Comp, 2 Inc, 0 Err │
│ Tok: 145.0 gen/s, 454.5 tot/s, 36.9ms TTFT, 13.7ms ITL, 546 Prompt, 256 Gen │
│ [18:13:57] ⠋ 100% concurrent@4 (complete) Req: 1.1 req/s, 3.59s Lat, 4.0 Conc, 64 Comp, 4 Inc, 0 Err │
│ Tok: 284.8 gen/s, 892.7 tot/s, 59.0ms TTFT, 13.9ms ITL, 546 Prompt, 256 Gen │
│ [18:15:02] ⠋ 100% concurrent@8 (complete) Req: 2.2 req/s, 3.70s Lat, 8.0 Conc, 128 Comp, 7 Inc, 0 Err │
│ Tok: 553.5 gen/s, 1735.2 tot/s, 79.8ms TTFT, 14.2ms ITL, 547 Prompt, 256 Gen │
│ [18:16:08] ⠋ 100% concurrent@16 (complete) Req: 4.2 req/s, 3.83s Lat, 16.0 Conc, 240 Comp, 16 Inc, 0 Err │
│ Tok: 1066.9 gen/s, 3344.6 tot/s, 97.5ms TTFT, 14.6ms ITL, 547 Prompt, 256 Gen │
│ [18:17:13] ⠋ 100% concurrent@32 (complete) Req: 8.1 req/s, 3.94s Lat, 31.8 Conc, 480 Comp, 31 Inc, 0 Err │
│ Tok: 2069.7 gen/s, 6488.4 tot/s, 120.8ms TTFT, 15.0ms ITL, 547 Prompt, 256 Gen │
│ [18:18:20] ⠋ 100% concurrent@64 (complete) Req: 13.6 req/s, 4.60s Lat, 62.3 Conc, 813 Comp, 57 Inc, 0 Err │
│ Tok: 3472.1 gen/s, 10884.9 tot/s, 190.9ms TTFT, 17.3ms ITL, 547 Prompt, 256 Gen │
│ [18:19:28] ⠋ 100% concurrent@128 (complete) Req: 16.8 req/s, 7.37s Lat, 123.5 Conc, 1005 Comp, 126 Inc, 0 Err │
│ Tok: 4289.1 gen/s, 13445.8 tot/s, 356.4ms TTFT, 27.5ms ITL, 547 Prompt, 256 Gen │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
Generating... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ (8/8) [ 0:08:43 < 0:00:00 ]
Benchmarks Metadata:
Run id:8ccb6da1-83f4-4624-8d84-07c723b0b2a5
Duration:530.4 seconds
Profile:type=concurrent, strategies=['concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent', 'concurrent'], streams=[1, 2, 4, 8, 16, 32, 64, 128]
Args:max_number=None, max_duration=60.0, warmup_number=None, warmup_duration=3.0, cooldown_number=None, cooldown_duration=None
Worker:type_='generative_requests_worker' backend_type='openai_http' backend_target='http://vllm-server:8000' backend_model='meta-llama/Llama-3.2-3B-Instruct' backend_info={'max_output_tokens':
16384, 'timeout': 300, 'http2': True, 'follow_redirects': True, 'headers': {}, 'text_completions_path': '/v1/completions', 'chat_completions_path': '/v1/chat/completions'}
Request Loader:type_='generative_request_loader' data='prompt_tokens=512,output_tokens=256' data_args=None processor='meta-llama/Llama-3.2-3B-Instruct' processor_args=None
Extras:None
Benchmarks Info:
=====================================================================================================================================================
Metadata |||| Requests Made ||| Prompt Tok/Req ||| Output Tok/Req ||| Prompt Tok Total||| Output Tok Total ||
Benchmark| Start Time| End Time| Duration (s)| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err| Comp| Inc| Err
--------------|-----------|---------|-------------|------|-----|-----|------|------|----|------|------|----|-------|------|----|--------|------|-----
concurrent@1| 18:11:52| 18:12:52| 60.0| 17| 1| 0| 546.5| 512.0| 0.0| 256.0| 231.0| 0.0| 9291| 512| 0| 4352| 231| 0
concurrent@2| 18:12:57| 18:13:57| 60.0| 32| 2| 0| 546.5| 512.0| 0.0| 256.0| 251.0| 0.0| 17488| 1024| 0| 8192| 502| 0
concurrent@4| 18:14:02| 18:15:02| 60.0| 64| 4| 0| 546.4| 512.0| 0.0| 256.0| 175.2| 0.0| 34972| 2048| 0| 16384| 701| 0
concurrent@8| 18:15:07| 18:16:07| 60.0| 128| 7| 0| 546.6| 512.0| 0.0| 256.0| 50.7| 0.0| 69966| 3584| 0| 32768| 355| 0
concurrent@16| 18:16:13| 18:17:13| 60.0| 240| 16| 0| 546.5| 512.0| 0.0| 256.0| 166.0| 0.0| 131170| 8192| 0| 61440| 2656| 0
concurrent@32| 18:17:18| 18:18:18| 60.0| 480| 31| 0| 546.5| 512.0| 0.0| 256.0| 47.4| 0.0| 262339| 15872| 0| 122880| 1468| 0
concurrent@64| 18:18:25| 18:19:25| 60.0| 813| 57| 0| 546.5| 512.0| 0.0| 256.0| 110.7| 0.0| 444341| 29184| 0| 208128| 6311| 0
concurrent@128| 18:19:33| 18:20:33| 60.0| 1005| 126| 0| 546.5| 512.0| 0.0| 256.0| 65.8| 0.0| 549264| 64512| 0| 257280| 8296| 0
=====================================================================================================================================================
Benchmarks Stats:
=======================================================================================================================================================
Metadata | Request Stats || Out Tok/sec| Tot Tok/sec| Req Latency (sec) ||| TTFT (ms) ||| ITL (ms) ||| TPOT (ms) ||
Benchmark| Per Second| Concurrency| mean| mean| mean| median| p99| mean| median| p99| mean| median| p99| mean| median| p99
--------------|-----------|------------|------------|------------|------|--------|------|------|-------|-------|-----|-------|-----|-----|-------|-----
concurrent@1| 0.30| 1.00| 76.4| 239.4| 3.35| 3.35| 3.38| 29.6| 29.0| 38.9| 13.0| 13.0| 13.1| 13.0| 13.0| 13.0
concurrent@2| 0.57| 2.00| 145.0| 454.5| 3.53| 3.53| 3.55| 36.9| 39.0| 59.6| 13.7| 13.7| 13.8| 13.6| 13.7| 13.7
concurrent@4| 1.11| 4.00| 284.8| 892.7| 3.59| 3.59| 3.65| 59.0| 65.7| 88.2| 13.9| 13.8| 14.1| 13.8| 13.8| 14.0
concurrent@8| 2.16| 7.99| 553.5| 1735.2| 3.70| 3.69| 3.76| 79.8| 80.7| 152.6| 14.2| 14.2| 14.5| 14.1| 14.1| 14.4
concurrent@16| 4.17| 15.97| 1066.9| 3344.6| 3.83| 3.82| 3.99| 97.5| 96.3| 283.9| 14.6| 14.6| 14.9| 14.6| 14.6| 14.8
concurrent@32| 8.08| 31.84| 2069.7| 6488.4| 3.94| 3.90| 4.31| 120.8| 101.7| 564.3| 15.0| 14.9| 15.9| 14.9| 14.8| 15.9
concurrent@64| 13.56| 62.34| 3472.1| 10884.9| 4.60| 4.54| 5.43| 190.9| 133.9| 1113.2| 17.3| 17.2| 18.2| 17.2| 17.2| 18.2
concurrent@128| 16.75| 123.45| 4289.1| 13445.8| 7.37| 7.21| 9.21| 356.4| 161.9| 2319.9| 27.5| 27.5| 28.8| 27.4| 27.4| 28.7
=======================================================================================================================================================
Saving benchmarks report...
Benchmarks report saved to /benchmarks.json
Benchmarking complete.

BIN
benchmarking/k8s-benchmark/results/vllm_replica1_benchmark_results.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 562 KiB