Posted on original issue

Alright, here are the results with the lowest memory (128m) using the same 99 metrics -- @tgip-work @nmoutschen @michaelbrewer

I'm pretty happy with it, and will not proceed with ujson as the underlying machine Lambda uses doesn't seem to make much use tbh - it could be different in other scenarios perhaps. I'll adjust the test SLA to take into account GitHub CI slower machines but I'd consider it solved... let me know otherwise @tgip-work :)

BEFORE - Current public Powertools

Fastest Lambda execution according to X-Ray: 54ms

All virtual users finished
Summary report @ 16:09:49(+0100) 2021-03-04
Scenarios launched: 10
Scenarios completed: 10
Requests completed: 2000
Mean response/sec: 48.23
Response time (msec):
min: 124.1
max: 1212.6
median: 189.9
p95: 244.5
p99: 385.2
Scenario counts:
0: 10 (100%)
Codes:
200: 1999
502: 1

AFTER -- New Powertools with Lazy Loading and this PR optimization

Fastest Lambda execution according to X-Ray: 2.2ms

All virtual users finished
Summary report @ 16:13:18(+0100) 2021-03-04
Scenarios launched: 10
Scenarios completed: 10
Requests completed: 2000
Mean response/sec: 100.5
Response time (msec):
min: 63.1
max: 1128.9
median: 81.1
p95: 132.5
p99: 209
Scenario counts:
0: 10 (100%)
Codes:
200: 2000

AFTER - New optimizations + ujson

Fastest Lambda execution according to X-Ray: 2.6ms

All virtual users finished
Summary report @ 16:28:12(+0100) 2021-03-04
Scenarios launched: 10
Scenarios completed: 10
Requests completed: 2000
Mean response/sec: 105.71
Response time (msec):
min: 62.3
max: 525.1
median: 81
p95: 134
p99: 211.5
Scenario counts:
0: 10 (100%)
Codes:
200: 2000

Details

Load test with artillery using REST API + Lambda (could be even faster w/ HTTP API): artillery quick --count 10 -n 200 <endpoint>

Source code

from aws_lambda_powertools import Metrics

# https://awslabs.github.io/aws-lambda-powertools-python/#features
metrics = Metrics(namespace="perftest", service="perftest")


@metrics.log_metrics
def lambda_handler(event, context):
    for i in range(99):
        metrics.add_metric(name=f"metric_{i}", value=1, unit="Count")

    return {"statusCode": 200, "body": "hello world"}

Codecov Report

Merging #307 (fcdbeb8) into develop (b4d0baa) will decrease coverage by 0.24%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop     #307      +/-   ##
===========================================
- Coverage    99.78%   99.54%   -0.25%     
===========================================
  Files           91       90       -1     
  Lines         3315     3291      -24     
  Branches       157      160       +3     
===========================================
- Hits          3308     3276      -32     
- Misses           5       10       +5     
- Partials         2        5       +3     

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b4d0baa...fcdbeb8. Read the comment docs.

thanks @nmoutschen ;) 🎉