Ask: reason for poor performance, http echo

[sourcey]

Ask: reason for poor performance, http echo

[auscaster]

Yes that should be all you need to compile correctly, although it does seem slow. Can you please add VERBOSE=1 to the CMake commands and send the the full build command? Also what architecture are you running? What results do you get from a basic nodejs echo server?

I just tried it myself on an ubuntu 16.10 virtualbox installation with 4 cores and got the following, and the result was pretty damn fast:

wrk -d10s --timeout 2s http://localhost:1337
Running 10s test @ http://localhost:1337
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   430.91us  521.60us  13.94ms   98.29%
    Req/Sec    12.95k     1.55k   15.63k    70.00%
  257774 requests in 10.01s, 24.34MB read
Requests/sec:  25762.27
Transfer/sec:      2.43MB

[diorahman]

This is the log: log.txt

Also what architecture are you running?

$ dpkg --print-architecture
amd64
$ lscpu
Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                4
On-line CPU(s) list:   0-3
Thread(s) per core:    2
Core(s) per socket:    2
Socket(s):             1
NUMA node(s):          1
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 78
Model name:            Intel(R) Core(TM) i7-6560U CPU @ 2.20GHz
Stepping:              3
CPU MHz:               499.914
CPU max MHz:           3200.0000
CPU min MHz:           400.0000
BogoMIPS:              4415.86
Virtualization:        VT-x
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K
L3 cache:              4096K
NUMA node0 CPU(s):     0-3
Flags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch epb intel_pt tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid mpx rdseed adx smap clflushopt xsaveopt xsavec xgetbv1 dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp

What results do you get from a basic nodejs echo server?

With following code:

var http = require('http');
var server = http.createServer(function (request, response) {
  response.writeHead(200, {"Content-Type": "text/plain"});
  response.end("Hello World\n");
});
server.listen(8000);

$ ./wrk -d10s --timeout 2s http://localhost:8000
Running 10s test @ http://localhost:8000
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   376.71us  454.50us  16.41ms   99.05%
    Req/Sec    14.37k     1.25k   14.97k    96.53%
  288938 requests in 10.10s, 42.99MB read
Requests/sec:  28608.71
Transfer/sec:      4.26MB

[auscaster]

Thanks for your response. The build certainly seems to be in order, and there is no apparent reason for the slowness on your architecture. Sorry I can't be of more help right this moment, but I will certainly be running some more tests on other boxes and profiling when I have a spare few hours.

[auscaster]

Hello there, since your bug request I've been looking deep into the cause of your performance issues and it looks like I have found the root. The garbage collection interval was slightly too large which was choking the event loop under a huge connection load when too much memory was being freed at once.

I've made a number of other performance updates to both the net and http modules and the result is the echo server is consistently performing almost twice as fast as nodejs.

$ wrk -d10s --timeout 2s http://localhost:1337
Running 10s test @ http://localhost:1337
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   265.76us  472.62us  12.42ms   97.51%
    Req/Sec    18.84k     1.26k   21.57k    74.50%
  375060 requests in 10.01s, 20.39MB read
Requests/sec:  37461.50
Transfer/sec:      2.04MB

I would be very interested to hear your thoughts and findings. Thanks again :)