Benchmark for latency and throughput both GPU server and Modal#12
Closed
ArnavBharti wants to merge 14 commits intomainfrom
Closed
Benchmark for latency and throughput both GPU server and Modal#12ArnavBharti wants to merge 14 commits intomainfrom
ArnavBharti wants to merge 14 commits intomainfrom
Conversation
commented out os.environ["CUDA_VISIBLE_DEVICES"] = "0" line.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
=== Benchmark Summary ===
Mode: server
Target: https://scaledfocus--masala-embed-server-modal-fastapi-app.modal.run/v1/embeddings
Concurrency: 100
Payloads: 3000
Request timeout: 120s
=== Per-stage results ===
--- stage-1 (duration=20s, target_rps=5) ---
Requests: 100 Successes: 100 Failures: 0
Measured wall-time (first->last): 16.251 s
Throughput (successful reqs / wall-time): 6.154 req/s
End-to-end latency p50/p95/p99: 475.93 / 3199.36 / 4002.60 ms
vLLM p50/p95/p99: 166.60 / 344.90 / 364.09 ms
Normalize p50/p95/p99: 0.01 / 0.02 / 0.04 ms
Search p50/p95/p99: 0.17 / 0.26 / 0.29 ms
Server p50/p95/p99: 166.90 / 345.12 / 364.27 ms
--- stage-2 (duration=40s, target_rps=20) ---
Requests: 796 Successes: 796 Failures: 0
Measured wall-time (first->last): 40.084 s
Throughput (successful reqs / wall-time): 19.858 req/s
End-to-end latency p50/p95/p99: 513.94 / 588.97 / 633.42 ms
vLLM p50/p95/p99: 191.36 / 251.17 / 282.14 ms
Normalize p50/p95/p99: 0.01 / 0.02 / 0.02 ms
Search p50/p95/p99: 0.14 / 0.25 / 0.40 ms
Server p50/p95/p99: 191.60 / 251.36 / 282.32 ms
--- stage-3 (duration=60s, target_rps=50) ---
Requests: 2980 Successes: 2980 Failures: 0
Measured wall-time (first->last): 60.127 s
Throughput (successful reqs / wall-time): 49.562 req/s
End-to-end latency p50/p95/p99: 656.43 / 877.53 / 1108.32 ms
vLLM p50/p95/p99: 295.64 / 448.96 / 521.56 ms
Normalize p50/p95/p99: 0.01 / 0.02 / 0.04 ms
Search p50/p95/p99: 0.17 / 0.42 / 0.61 ms
Server p50/p95/p99: 295.85 / 449.30 / 521.91 ms
--- stage-4 (duration=60s, target_rps=100) ---
Requests: 5960 Successes: 5960 Failures: 0
Measured wall-time (first->last): 90.897 s
Throughput (successful reqs / wall-time): 65.569 req/s
End-to-end latency p50/p95/p99: 1513.67 / 1903.91 / 2114.91 ms
vLLM p50/p95/p99: 314.50 / 891.68 / 1387.54 ms
Normalize p50/p95/p99: 0.01 / 0.02 / 0.14 ms
Search p50/p95/p99: 0.20 / 0.60 / 0.78 ms
Server p50/p95/p99: 314.76 / 891.96 / 1387.78 ms
=== Overall ===
Requests: 9836 Successes: 9836 Failures: 0
Measured wall-time (first->last): 208.773 s
Throughput (successful reqs / wall-time): 47.113 req/s
End-to-end latency p50/p95/p99: 1330.15 / 1823.62 / 2086.31 ms
vLLM p50/p95/p99: 295.48 / 685.89 / 1349.32 ms
Normalize p50/p95/p99: 0.01 / 0.02 / 0.10 ms
Search p50/p95/p99: 0.18 / 0.55 / 0.74 ms
Server p50/p95/p99: 295.81 / 686.11 / 1349.63 ms