Hugging Face's logo

Datasets:
kshitijthakkar
/
mcp-server-bench-gradio-optimized

Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
mcp
benchmark
gradio
fastmcp
api-performance
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
kshitijthakkar commited on
Commit
e14ae2d
·
verified ·
1 Parent(s): fb79045

Update benchmark report, charts, and detailed results

Browse files
Files changed (6) hide show
  1. README.md +119 -61
  2. detailed_results.json +0 -0
  3. plots/gradio_cl_scaling.png +3 -0
  4. plots/latency_comparison.png +3 -0
  5. plots/protocol_overhead.png +3 -0
  6. plots/throughput_comparison.png +3 -0
README.md CHANGED
@@ -1,63 +1,121 @@
1
  ---
2
- dataset_info:
3
- features:
4
- - name: scenario_id
5
- dtype: large_string
6
- - name: server
7
- dtype: large_string
8
- - name: protocol
9
- dtype: large_string
10
- - name: tool
11
- dtype: large_string
12
- - name: virtual_users
13
- dtype: int64
14
- - name: concurrency_limit
15
- dtype: float64
16
- - name: duration_s
17
- dtype: float64
18
- - name: total_requests
19
- dtype: int64
20
- - name: successful_requests
21
- dtype: int64
22
- - name: failed_requests
23
- dtype: int64
24
- - name: error_rate_pct
25
- dtype: float64
26
- - name: throughput_rps
27
- dtype: float64
28
- - name: latency_min_ms
29
- dtype: float64
30
- - name: latency_p50_ms
31
- dtype: float64
32
- - name: latency_p75_ms
33
- dtype: float64
34
- - name: latency_p90_ms
35
- dtype: float64
36
- - name: latency_p95_ms
37
- dtype: float64
38
- - name: latency_p99_ms
39
- dtype: float64
40
- - name: latency_max_ms
41
- dtype: float64
42
- - name: latency_mean_ms
43
- dtype: float64
44
- - name: latency_stddev_ms
45
- dtype: float64
46
- - name: avg_cpu_pct
47
- dtype: float64
48
- - name: avg_memory_mb
49
- dtype: float64
50
- - name: peak_memory_mb
51
- dtype: float64
52
- splits:
53
- - name: train
54
- num_bytes: 12854
55
- num_examples: 48
56
- download_size: 16712
57
- dataset_size: 12854
58
- configs:
59
- - config_name: default
60
- data_files:
61
- - split: train
62
- path: data/train-*
63
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ language:
3
+ - en
4
+ license: mit
5
+ tags:
6
+ - mcp
7
+ - benchmark
8
+ - gradio
9
+ - fastmcp
10
+ - api-performance
11
+ pretty_name: MCP Server Benchmark Results
12
+ size_categories:
13
+ - n<1K
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  ---
15
+
16
+ # 🔬 Gradio vs FastMCP Benchmark Report
17
+
18
+ **Generated:** 2026-03-02T13:04:10.857215
19
+
20
+ **Total scenarios:** 48
21
+
22
+ ## Executive Summary
23
+
24
+ - **echo**: Fastmcp wins (96.6 vs 176.5 RPS, 1.83x difference)
25
+ - Gradio best config: concurrency_limit=nan
26
+ - **fibonacci**: Fastmcp wins (43.5 vs 57.1 RPS, 1.31x difference)
27
+ - Gradio best config: concurrency_limit=nan
28
+ - **async_sleep**: Gradio wins (93.1 vs 80.2 RPS, 1.16x difference)
29
+ - Gradio best config: concurrency_limit=nan
30
+ - **payload_echo**: Fastmcp wins (82.5 vs 164.8 RPS, 2.0x difference)
31
+ - Gradio best config: concurrency_limit=nan
32
+
33
+ ## Throughput (Requests/Second)
34
+
35
+ | | ('fastmcp', 'http_api') | ('fastmcp', 'mcp_streamable') | ('gradio', 'http_api') | ('gradio', 'mcp_streamable') |
36
+ |:---------------------|--------------------------:|--------------------------------:|-------------------------:|-------------------------------:|
37
+ | ('async_sleep', 1) | 15.76 | 14.41 | 8.54 | 14.23 |
38
+ | ('async_sleep', 10) | 80.21 | 56.47 | 56.79 | 93.11 |
39
+ | ('async_sleep', 25) | 42.1 | 40.13 | 37.4 | 56.76 |
40
+ | ('echo', 1) | 176.5 | 72.37 | 11.16 | 72.12 |
41
+ | ('echo', 10) | 79.84 | 43.33 | 49.37 | 96.56 |
42
+ | ('echo', 25) | 62.5 | 39.44 | 37.87 | 70.43 |
43
+ | ('fibonacci', 1) | 45.39 | 35.94 | 11.91 | 35.18 |
44
+ | ('fibonacci', 10) | 57.08 | 47.93 | 37.39 | 43.5 |
45
+ | ('fibonacci', 25) | 51.19 | 35.94 | 32.67 | 43.21 |
46
+ | ('payload_echo', 1) | 164.84 | 80.1 | 11.62 | 70.56 |
47
+ | ('payload_echo', 10) | 91.25 | 50.53 | 49.93 | 82.54 |
48
+ | ('payload_echo', 25) | 66.84 | 41.57 | 37.2 | 61.16 |
49
+
50
+ ## Latency p50 (ms)
51
+
52
+ | | ('fastmcp', 'http_api') | ('fastmcp', 'mcp_streamable') | ('gradio', 'http_api') | ('gradio', 'mcp_streamable') |
53
+ |:---------------------|--------------------------:|--------------------------------:|-------------------------:|-------------------------------:|
54
+ | ('async_sleep', 1) | 62.599 | 64.32 | 112.296 | 75.659 |
55
+ | ('async_sleep', 10) | 105.196 | 159.981 | 155.612 | 102.057 |
56
+ | ('async_sleep', 25) | 430.916 | 528.933 | 546.454 | 337.852 |
57
+ | ('echo', 1) | 5.005 | 13.005 | 101.503 | 13.037 |
58
+ | ('echo', 10) | 102.519 | 201.01 | 177.598 | 95.355 |
59
+ | ('echo', 25) | 298.811 | 529.175 | 549.508 | 275.746 |
60
+ | ('fibonacci', 1) | 20.998 | 26.535 | 66.875 | 27.521 |
61
+ | ('fibonacci', 10) | 166.862 | 189.789 | 266.299 | 226.342 |
62
+ | ('fibonacci', 25) | 463.179 | 591.419 | 639.153 | 573.479 |
63
+ | ('payload_echo', 1) | 5.998 | 12 | 81.067 | 13.662 |
64
+ | ('payload_echo', 10) | 95.531 | 175.658 | 175.569 | 110.116 |
65
+ | ('payload_echo', 25) | 336.98 | 514.13 | 557.964 | 321.014 |
66
+
67
+ ## Gradio concurrency_limit Scaling
68
+
69
+ How does Gradio's throughput change as concurrency_limit increases?
70
+
71
+
72
+
73
+ ## Protocol Overhead: HTTP API vs MCP
74
+
75
+ Comparing latency of the same tool called via REST API vs MCP protocol:
76
+
77
+ | | http_api | mcp_streamable |
78
+ |:----------------------------|-----------:|-----------------:|
79
+ | ('fastmcp', 'async_sleep') | 199.57 | 251.078 |
80
+ | ('fastmcp', 'echo') | 135.445 | 247.73 |
81
+ | ('fastmcp', 'fibonacci') | 217.013 | 269.248 |
82
+ | ('fastmcp', 'payload_echo') | 146.17 | 233.929 |
83
+ | ('gradio', 'async_sleep') | 271.454 | 171.856 |
84
+ | ('gradio', 'echo') | 276.203 | 128.046 |
85
+ | ('gradio', 'fibonacci') | 324.109 | 275.781 |
86
+ | ('gradio', 'payload_echo') | 271.533 | 148.264 |
87
+
88
+ ## Error Rates
89
+
90
+ All scenarios completed with 0% error rate. ✅
91
+
92
+
93
+ ## Resource Usage
94
+
95
+ | server | ('mean', 'avg_cpu_pct') | ('mean', 'avg_memory_mb') | ('mean', 'peak_memory_mb') | ('max', 'avg_cpu_pct') | ('max', 'avg_memory_mb') | ('max', 'peak_memory_mb') |
96
+ |:---------|--------------------------:|----------------------------:|-----------------------------:|-------------------------:|---------------------------:|----------------------------:|
97
+ | fastmcp | 0 | 6.14458 | 6.14458 | 0 | 6.15 | 6.15 |
98
+ | gradio | 0 | 6.14417 | 6.14417 | 0 | 6.15 | 6.15 |
99
+
100
+ ## Benchmark Charts
101
+
102
+ ![Throughput Comparison](plots/throughput_comparison.png)
103
+
104
+ ![Latency Comparison](plots/latency_comparison.png)
105
+
106
+ ![Gradio Concurrency Limit Scaling](plots/gradio_cl_scaling.png)
107
+
108
+ ![Protocol Overhead](plots/protocol_overhead.png)
109
+
110
+ ## Methodology
111
+
112
+ - Both servers use identical tool implementations (imported from shared_tools.py)
113
+ - Each scenario runs in an isolated server subprocess
114
+ - Warmup period excluded from measurements
115
+ - Load generated by async httpx workers (not external tools)
116
+ - MCP tests use full protocol lifecycle (initialize → call_tool)
117
+ - System metrics sampled every 1s via psutil
118
+
119
+ ---
120
+
121
+ *Benchmarks generated by [mcp-server-bench](https://github.com/Mandark-droid/mcp-server-bench)*
detailed_results.json ADDED
The diff for this file is too large to render. See raw diff
 
plots/gradio_cl_scaling.png ADDED

Git LFS Details

  • SHA256: 40686d7285953520e3262c98d47098c09a1d27a869684d6f16198a99b78cb7dc
  • Pointer size: 131 Bytes
  • Size of remote file: 142 kB
plots/latency_comparison.png ADDED

Git LFS Details

  • SHA256: 5aa6ac2aeaa331d559ac878414e04a4980ac0fc5880856bdafa542d7392a0f05
  • Pointer size: 130 Bytes
  • Size of remote file: 46.1 kB
plots/protocol_overhead.png ADDED

Git LFS Details

  • SHA256: 62f029ce26efbfd446026704b055fbe4217ead48c2fcf695e643f1babed702f8
  • Pointer size: 130 Bytes
  • Size of remote file: 48.9 kB
plots/throughput_comparison.png ADDED

Git LFS Details

  • SHA256: 300c460706577270819d97a70a89f21a5d34e0bb926f033c8a4f70117bb0dd50
  • Pointer size: 131 Bytes
  • Size of remote file: 179 kB