Using vegeta in a community driven benchmarking project

[waghanza]

Hi,

I'm leading a benchmarking project https://github.com/the-benchmarker/web-frameworks.

We are currently using wrk, but it seems not accurate to our goal.

I've starred in a thread to discuss changing our toolset, https://github.com/the-benchmarker/web-frameworks/discussions/8088.

Any insights of any community members here is very appreciated :heart:

[waghanza]

We actually use wrk (without rate limit) and I have some results that I can not understand.

Testing this line https://github.com/the-benchmarker/web-frameworks/blob/aedc5b0a39a18840b7818906f389f6277b66cbd1/rust/actix/src/main.rs#L12 with wrk, wrk http://172.17.0.6:3000

I have

Running 10s test @ http://172.17.0.6:3000
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    35.73us   73.57us   2.77ms   97.49%
    Req/Sec   167.36k    11.09k  180.08k    91.58%
  3364219 requests in 10.10s, 240.63MB read
Requests/sec: 333098.97
Transfer/sec:     23.83MB

and with vegeta, disabling rate limit (echo "GET http://172.17.0.6:3000" | ~/bin/vegeta attack -duration 10s -max-workers 2 -rate 0 -max-connections 10 | tee results.bin | vegeta report -type=json | python -m json.tool -) , I have

{
    "latencies": {
        "total": 19222679102,
        "mean": 80362,
        "50th": 80846,
        "90th": 89392,
        "95th": 99321,
        "99th": 160442,
        "max": 7349046,
        "min": 14917
    },
    "bytes_in": {
        "total": 0,
        "mean": 0
    },
    "bytes_out": {
        "total": 0,
        "mean": 0
    },
    "earliest": "2025-01-03T09:26:06.784554552+01:00",
    "latest": "2025-01-03T09:26:16.78456101+01:00",
    "end": "2025-01-03T09:26:16.784617968+01:00",
    "duration": 10000006458,
    "wait": 56958,
    "requests": 239201,
    "rate": 23920.084552409397,
    "throughput": 23919.948309255804,
    "success": 1,
    "status_codes": {
        "200": 239201
    },
    "errors": []
}

and with an other rust based https://github.com/the-benchmarker/web-frameworks/blob/aedc5b0a39a18840b7818906f389f6277b66cbd1/rust/axum/src/main.rs#L8

• vegeta

{
    "latencies": {
        "total": 29042425944,
        "mean": 82294,
        "50th": 81431,
        "90th": 89874,
        "95th": 97185,
        "99th": 169418,
        "max": 12406035,
        "min": 13542
    },
    "bytes_in": {
        "total": 0,
        "mean": 0
    },
    "bytes_out": {
        "total": 0,
        "mean": 0
    },
    "earliest": "2025-01-03T09:52:19.095929094+01:00",
    "latest": "2025-01-03T09:52:34.095824521+01:00",
    "end": "2025-01-03T09:52:34.095885729+01:00",
    "duration": 14999895427,
    "wait": 61208,
    "requests": 352907,
    "rate": 23527.297354671085,
    "throughput": 23527.20135047244,
    "success": 1,
    "status_codes": {
        "200": 352907
    },
    "errors": []
}

• wrk

Running 10s test @ http://172.17.0.7:3000
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   154.99us  322.14us  14.18ms   99.25%
    Req/Sec    36.47k     2.83k   65.80k    93.03%
  729313 requests in 10.10s, 52.16MB read
Requests/sec:  72209.68
Transfer/sec:      5.16MB

and with a nim based framework =>https://github.com/the-benchmarker/web-frameworks/blob/aedc5b0a39a18840b7818906f389f6277b66cbd1/nim/prologue/server.nim#L3

• vegeta

{
    "latencies": {
        "total": 29113548225,
        "mean": 94888,
        "50th": 87280,
        "90th": 109852,
        "95th": 130171,
        "99th": 259565,
        "max": 4068316,
        "min": 21292
    },
    "bytes_in": {
        "total": 0,
        "mean": 0
    },
    "bytes_out": {
        "total": 0,
        "mean": 0
    },
    "earliest": "2025-01-03T10:15:11.805718673+01:00",
    "latest": "2025-01-03T10:15:26.805640684+01:00",
    "end": "2025-01-03T10:15:26.80573285+01:00",
    "duration": 14999922011,
    "wait": 92166,
    "requests": 306818,
    "rate": 20454.639682459612,
    "throughput": 20454.514001090334,
    "success": 1,
    "status_codes": {
        "200": 306818
    },
    "errors": []
}

• wrk

Running 10s test @ http://172.17.0.10:3000
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    49.15us  131.55us   7.76ms   98.40%
    Req/Sec   114.19k    11.45k  207.48k    93.03%
  2283029 requests in 10.10s, 285.22MB read
Requests/sec: 226054.00
Transfer/sec:     28.24MB

Any idea on how to explain those figures (I mean the difference between those tools) ?

cc @appleboy @tsenart