PoC for external calls via HTTP/POST+JSON

[garypen]

DO NOT MERGE UNDER ANY CIRCUMSTANCES

PoC for external calls via HTTP/POST+JSON payloads via rhai extension method

I've done some fairly unscientific comparative benchmarking on my laptop to give a rough indication of the overhead of the current implementation.

Experiment: Send

• Context
• Headers
• Body
• Supergraph Schema
  to an "echo" server which does nothing to the payload but echo it back to the router.

hey was used for driving load. Configured as:

hey \
    -cpus 8 \
    -m POST \
    -n 120 \
    -z 60s \
    -t 60 \
    -c 100 \
    -H 'content-type:application/json' \
    -H 'accept:application/json' \
    -d '{"query":"query {\n  products(limit: 1) {\n upc \n name\n price \n }\n}\n","variables":{}}' \
    -h2 \
    'http://127.0.0.1:4000'

Without External:

Memory: 85MB
CPU: 484

./HEY.local

Summary:
Total: 60.0044 secs
Slowest: 0.1302 secs
Fastest: 0.0005 secs
Average: 0.0060 secs
Requests/sec: 24631.3440

Response time histogram:
0.001 [1] |
0.014 [996137] |■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
0.026 [3095] |
0.039 [509] |
0.052 [157] |
0.065 [41] |
0.078 [28] |
0.091 [18] |
0.104 [9] |
0.117 [4] |
0.130 [1] |

Latency distribution:
10% in 0.0026 secs
25% in 0.0031 secs
50% in 0.0037 secs
75% in 0.0046 secs
90% in 0.0057 secs
95% in 0.0066 secs
99% in 0.0095 secs

Details (average, fastest, slowest):
DNS+dialup: 0.0000 secs, 0.0005 secs, 0.1302 secs
DNS-lookup: 0.0000 secs, 0.0000 secs, 0.0000 secs
req write: 0.0000 secs, 0.0000 secs, 0.0083 secs
resp wait: 0.0059 secs, 0.0003 secs, 0.1301 secs
resp read: 0.0001 secs, 0.0000 secs, 0.0447 secs

Status code distribution:
[200] 1000000 responses

With External (echo to local server):

Memory: 95MB
CPU: 300

garypen@Garys-MBP TV2_load % ./HEY.local

Summary:
Total: 60.0057 secs
Slowest: 0.2101 secs
Fastest: 0.0017 secs
Average: 0.0113 secs
Requests/sec: 8838.8037

Response time histogram:
0.002 [1] |
0.023 [523684] |■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
0.043 [6530] |
0.064 [63] |
0.085 [1] |
0.106 [0] |
0.127 [0] |
0.148 [0] |
0.168 [0] |
0.189 [0] |
0.210 [100] |

Latency distribution:
10% in 0.0071 secs
25% in 0.0086 secs
50% in 0.0106 secs
75% in 0.0132 secs
90% in 0.0163 secs
95% in 0.0184 secs
99% in 0.0233 secs

Details (average, fastest, slowest):
DNS+dialup: 0.0000 secs, 0.0017 secs, 0.2101 secs
DNS-lookup: 0.0000 secs, 0.0000 secs, 0.0000 secs
req write: 0.0000 secs, 0.0000 secs, 0.0049 secs
resp wait: 0.0113 secs, 0.0016 secs, 0.2065 secs
resp read: 0.0000 secs, 0.0000 secs, 0.0100 secs

Status code distribution:
[200] 530379 responses

Summary: 1000000 Txns without external calls, 530379 with them.

Without: median response time: 0.014ms
With: median response time: 0.023ms

Which implies, that on average, the externalisation process adds about 1ms to the transaction time. That's a reasonable expectation of a lower bound on externalisation. Most external processes will do more than just echo back the payload...

[github-actions]

@garypen your pull request is missing a changelog!

[github-actions]

@garypen, please add a /NEXT_CHANGELOG.md entry for this PR to the appropriate section!

[garypen]

Closing since this was delivered via config rather than using rhai in #2229