Boom – HTTP(S) load generator

Hmmm. I like the output better than wrk and wrk2, however, it doesn't seem particularly efficient at generating high enough load in comparison to wrk:

    (trent@imac:ttys003) (Thu/09:54) .. (~s/gopath/bin)
    % time ./boom -cpus 8 -n 10000 http://panther:8080/plaintext 
    10000 / 10000 Booooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo! 100.00 % 

    Summary:
      Total:	8.4509 secs.
      Slowest:	1.1200 secs.
      Fastest:	0.0007 secs.
      Average:	0.0359 secs.
      Requests/sec:	1092.3152
      Total Data Received:	138465 bytes.
      Response Size per Request:	15 bytes.

    Status code distribution:
      [200]	9231 responses

    Response time histogram:
      0.001 [1]	|
      0.113 [9009]	|∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
      0.225 [0]	|
      0.336 [0]	|
      0.448 [0]	|
      0.560 [0]	|
      0.672 [0]	|
      0.784 [0]	|
      0.896 [0]	|
      1.008 [39]	|
      1.120 [182]	|

    Latency distribution:
      10% in 0.0023 secs.
      25% in 0.0053 secs.
      50% in 0.0100 secs.
      75% in 0.0161 secs.
      90% in 0.0255 secs.
      95% in 0.0343 secs.
      99% in 1.0158 secs.

    Error distribution:
      [769]	Get http://192.168.1.244:8080/plaintext: dial tcp 192.168.1.244:8080: can't assign requested address
    ./boom -cpus 8 -n 10000 http://panther:8080/plaintext  0.88s user 2.40s system 38% cpu 8.463 total

Ok, so, that took ~9 seconds to run 10,000 requests. Let's run wrk for 9 seconds and compare:

    (trent@imac:ttys003) (Thu/09:55) .. (~s/wrk)
    % ./wrk --latency -d 9 -t 8 -c 8 http://panther:8080/plaintext
    Running 9s test @ http://panther:8080/plaintext
      8 threads and 8 connections
      Thread Stats   Avg      Stdev     Max   +/- Stdev
        Latency   459.03us  263.24us   6.84ms   90.59%
        Req/Sec     2.27k   285.87     3.22k    76.54%
      Latency Distribution
         50%  407.00us
         75%  520.00us
         90%  675.00us
         99%    1.43ms
      153940 requests in 9.00s, 27.60MB read
    Requests/sec:  17111.78
    Transfer/sec:      3.07MB

That served 15x more requests?

Looks closer to siege[1] than ab.

https://www.joedog.org/siege-home/

Best bit of the README:

> Speed index: Hahahaha

How do you handle the case where the load generator runs out of resources (ie, file handlers, tcp sockets, network interface throughput, etc.) before it is able to generate sufficient load to test the server?