re: https://txt.black/~jack/we-got-it-all-wrong.txt

benchmarking back-to-back vs http
(https://github.com/jackdoe/back-to-back - sync i/o queue)

i made this very simplified benchmark to see the difference between
pull and push, in cpu intense and resource constrained environment.
(code: https://github.com/jackdoe/back-to-back/tree/master/sim)


simplified scenario back-to-back:

[ C1 ]
  |
[ B1 ]
  |
[ P1 ]


simplified scenario HTTP:

[ C1 ]
  |
[ P1 ]

C1 - consumer
P1 - producer (client)
B1 - broker


requests spin for random(100) time:
func handle() {
	for i := 0; i < rand.Intn(100); i++ {
		sha256.Sum256(make([]byte, 100000))
	}
}

initial results look very promising, the distribution of times from
back-to-back is very stable.


2019/02/19 21:33:37 btb ... 20000 messages, took: 34.21s, speed: 584.58 per second
-   0+10=   35   0.18% 
-  10+ 5=   20   0.27% 
-  15+ 8=  846   4.50% **
-  23+12=10708  58.05% ******************************
-  35+18= 8202  99.06% **********************
-  53+27=  174  99.92% 
-  80+40=   15 100.00% 
- 120+60=    0 100.00%
- 180+90=    0 100.00%
- 270+ 0=    0 100.00%

2019/02/19 21:34:12 http ... 20000 messages, took: 35.20s, speed: 568.22 per second
-   0+10=1638   8.19% *********
-  10+ 5=1554  15.96% ********
-  15+ 8=3123  31.57% *****************
-  23+12=4818  55.66% **************************
-  35+18=5407  82.70% ******************************
-  53+27=2936  97.38% ****************
-  80+40= 515  99.95% **
- 120+60=   9 100.00% 
- 180+90=   0 100.00%
- 270+ 0=   0 100.00%


run with one fast and one slow endpoint:

2019/02/19 22:26:27 btb ... 20000 messages, took: 29.95s, speed: 667.74 per second
-   0+10=9711  48.55% ******************************
-  10+ 5= 276  49.94% 
-  15+ 8=  20  50.03% 
-  23+12=  16  50.11% 
-  35+18=3398  67.11% **********
-  53+27=6549  99.85% ********************
-  80+40=  30 100.00% 
- 120+60=   0 100.00%
- 180+90=   0 100.00%
- 270+ 0=   0 100.00%

2019/02/19 22:26:56 http ... 20000 messages, took: 29.55s, speed: 676.91 per second
-   0+10=3274  16.37% ********************
-  10+ 5=1523  23.98% *********
-  15+ 8=4439  46.18% ****************************
-  23+12=4756  69.96% ******************************
-  35+18=3318  86.55% ********************
-  53+27=2037  96.73% ************
-  80+40= 581  99.64% ***
- 120+60=  70  99.99% 
- 180+90=   2 100.00% 
- 270+ 0=   0 100.00%




i believe that this will be even more obvious in multi hop environment
like modern micro services mesh, if 10% of the requests are slower,
in 3 hop env 27% (1 - 0.9**3) of the requests will be slower.

more test to be done

1) multi hop (micro services example)
2) packet loss
   (test with many consumers and one of them has some networking
   issues)

---
github.com/jackdoe Tue 19 Feb 22:12:43 CET 2019