Benchmarking HTTP/3

[bryancall]

/opt/bin/h2load --npn-list h3-29 -T 1 -N 1 -t 30 -n 200000 -c 100 cat urls.http3.config | xargs

finished in 35.49s, 5478.69 req/s, 86.47MB/s
requests: 200000 total, 194443 started, 194443 done, 194443 succeeded, 5557 failed, 5557 errored, 0 timeout
status codes: 194443 2xx, 0 3xx, 0 4xx, 0 5xx
traffic: 3.00GB (3217838371) total, 29.11MB (30527551) headers (space savings 35.12%), 2.97GB (3185754112) data
UDP datagram: 1573268 sent, 2947486 received
                     min         max         mean         sd        +/- sd
time for request:      502us    191.48ms     14.37ms      8.09ms    80.25%
time for connect:    14.40ms     18.41ms     16.01ms       893us    72.16%
time to 1st byte:    19.85ms     34.89ms     27.16ms      4.24ms    58.76%
req/s           :      58.75       81.37       70.11        4.65    65.98%

dstat 10 output:

----total-usage---- -dsk/total- -net/total- ---paging-- ---system--
usr sys idl wai stl| read  writ| recv  send|  in   out | int   csw
 98   1   0   0   0|  99k   25M|2406B 7209B|   0  6448 | 168k   31k
 98   2   0   0   0|  47k   25M|1320B 7445B|0.30  6318 | 165k   32k
 98   1   0   0   0| 112k   25M|1843B   17k|42.2  6451 | 170k   31k
 98   2   0   0   0|  83k   28M|2116B 7856B|16.6  7158 | 184k   30k
 98   1   0   0   0|  40k   30M|4038B 7533B|0.20  7655 | 192k   29k
 98   2   0   0   0|1848k   37M|1142B 7062B|2.30  9348 | 224k   30k
 99   1   0   0   0|  52k   17M|1478B 7099B|2.70  4291 | 124k   29k
 99   1   0   0   0|9422B   17M|2191B 5250B|0.30  4294 | 124k   29k
 98   1   0   0   0|  14k   25M|1284B 8480B|0.30  6339 | 165k   28k
 99   1   0   0   0|9831B   25M|1152B 7207B|   0  6414 | 167k   27k
 99   1   0   0   0|4506B   25M|4276B 7126B|0.30  6308 | 165k   26k
 98   2   0   0   0|8037k   41M|2187B 8874B|1563  10.7k| 252k   27k
 99   1   0   0   0|  45k   16M|1568B 6958B|   0  4071 | 120k   27k
 99   1   0   0   0| 819B   17M|1230B 6828B|0.20  4223 | 123k   26k
 99   1   0   0   0|  56k   25M|1208B   11k|1.30  6220 | 164k   26k
 99   1   0   0   0|  19k   19M|2298B 6909B|0.20  4839 | 136k   26k
 99   1   0   0   0|  15k   22M|3827B 8792B|0.40  5485 | 148k   25k

top:

top - 14:44:13 up  2:20,  1 user,  load average: 24.49, 22.60, 14.84
Tasks: 412 total,   1 running, 410 sleeping,   1 stopped,   0 zombie
%Cpu(s): 98.2 us,  1.1 sy,  0.0 ni,  0.1 id,  0.0 wa,  0.4 hi,  0.2 si,  0.0 st
MiB Mem :  64239.3 total,    433.3 free,  63249.6 used,    556.4 buff/cache
MiB Swap:  40432.0 total,  19131.4 free,  21300.5 used.    324.5 avail Mem

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
  30970 bcall     20   0   94.1g  54.0g   5588 S  2363  86.0 224:09.83 [TS_MAIN]. <-- RES == bad
  32075 bcall     20   0  454756  46904   3880 S  15.0   0.1   2:51.36 h2load
  16006 bcall     20   0 1457132 198792  15120 S   2.0   0.3   2:03.35 node
   3529 root      20   0   64.2g   2.9g   3016 S   0.7   4.7   0:44.36 java
  32583 bcall     20   0  241024   7392   5816 R   0.7   0.0   0:01.35 top
     78 root      rt   0       0      0      0 S   0.3   0.0   0:00.46 migration/13
   1251 mysql     20   0 1280096  49256   5732 S   0.3   0.1   0:00.43 mysqld

perf report:

# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 950K of event 'cycles'
# Event count (approx.): 953973010629
#
#   Overhead  Shared Object         Symbol                                              IPC   [IPC Coverage]
# ..........  ....................  ..................................................  ....................
#
      23.55%  traffic_server        [.] HQTransaction::get_transaction_id               -      -
      12.22%  traffic_server        [.] QUICStreamVCAdapter::unread_len                 -      -
       8.81%  traffic_server        [.] QUICStreamManager::will_generate_frame          -      -
       7.63%  traffic_server        [.] QUICBidirectionalStream::will_generate_frame    -      -
       6.98%  traffic_server        [.] QUICBidirectionalStream::generate_frame         -      -
       5.91%  traffic_server        [.] QUICFrameRetransmitter::create_retransmitted_f  -      -
       4.47%  traffic_server        [.] QUICFlowController::generate_frame              -      -
       3.50%  traffic_server        [.] QUICFlowController::will_generate_frame         -      -
       3.49%  traffic_server        [.] QUICStream::id                                  -      -
       2.86%  traffic_server        [.] QUICStreamManager::_handle_frame                -      -
       1.55%  traffic_server        [.] IOBufferReader::skip_empty_blocks               -      -
       1.48%  traffic_server        [.] QUICBidirectionalStreamStateMachine::is_allowe  -      -
       1.40%  traffic_server        [.] QUICStreamManager::total_offset_received        -      -
       1.29%  libpthread-2.33.so    [.] __pthread_mutex_lock                            -      -
       1.12%  traffic_server        [.] HQSession::get_transaction                      -      -
       1.00%  traffic_server        [.] QUICStreamManager::total_reordered_bytes        -      -
       0.80%  traffic_server        [.] QUICStreamManager::generate_frame               -      -
       0.68%  libpthread-2.33.so    [.] __pthread_getspecific                           -      -
       0.59%  traffic_server        [.] IOBufferReader::block_read_avail                -      -
       0.50%  [kernel.kallsyms]     [k] native_flush_tlb_local                          -      -
       0.45%  [kernel.kallsyms]     [k] __x86_indirect_thunk_rax                        -      -
       0.35%  traffic_server        [.] QUICFrameGenerator::_is_level_matched           -      -

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. Marking it stale to flag it for further consideration by the community.