Simulating Real Users

Simulating Real Users

Normal benchmarking just blasts requests as fast as possible. But real users don't work that way — they click a link, read the page, click another link, wait a bit, and so on.

User simulation mode makes your test behave like real people browsing your site.

How it works

Each "virtual user":

1. Sends a request
2. Waits for the response
3. Pauses for a bit (this is called "think time" — like a person reading the page)
4. Sends another request
5. Repeats until the test ends

Basic usage

# Simulate 100 users browsing for 60 seconds
pywrkr http://localhost:8080/ -u 100 -d 60

The -u flag switches pywrkr from "blast mode" to "user simulation mode."

Add think time

By default, each user waits 1 second between requests. You can change this:

# Users wait 2 seconds between requests (slow browsing)
pywrkr http://localhost:8080/ -u 100 -d 60 --think-time 2.0

# Users wait 0.5 seconds (fast clicking)
pywrkr http://localhost:8080/ -u 100 -d 60 --think-time 0.5

# No waiting at all (back to stress test, but with user count)
pywrkr http://localhost:8080/ -u 100 -d 60 --think-time 0

Gradual ramp-up

In real life, users don't all arrive at the same second. Use --ramp-up to start users gradually:

# 500 users, but start them over 30 seconds (not all at once)
pywrkr http://localhost:8080/ -u 500 -d 120 --ramp-up 30

This means:

• At second 0: ~0 users active
• At second 15: ~250 users active
• At second 30: all 500 users active
• Seconds 30-120: all 500 users running

Why ramp up? Starting all users at once creates a "thundering herd" — a huge spike that can crash your server before it even warms up. Ramp-up is more realistic and gives your server time to scale up (auto-scaling, connection pools, etc).

Add jitter to think time

Real users don't all wait exactly the same amount of time. Jitter adds randomness:

# Think time: 1 second +/- 50% (so 0.5 to 1.5 seconds randomly)
pywrkr http://localhost:8080/ -u 200 -d 60 --think-time 1.0 --think-jitter 0.5

The default jitter is 0.5 (50%), which is usually fine. Set it to 0 for perfectly uniform think time.

Full realistic example

Here's a test that closely simulates real traffic:

# 300 users, 5-minute test, 30-second ramp-up, 1.5s think time
pywrkr http://localhost:8080/ \
    -u 300 \
    -d 300 \
    --ramp-up 30 \
    --think-time 1.5 \
    --think-jitter 0.5

Reading the results

User simulation results include extra info:

  Mode:              300 virtual users, 300.0s
  Virtual Users:     300
  Ramp-up:           30.00s
  Think Time:        1.50s (+/-50%)
  Avg Reqs/User:     120.5

Avg Reqs/User tells you how many pages each virtual user visited during the test.

When to use user simulation vs. regular mode

Scenario	Use this
"How many requests/sec can my server handle?"	Regular mode: pywrkr url -c 100 -d 30
"Can my server handle 500 concurrent users?"	User simulation: pywrkr url -u 500 -d 60
"What happens during a traffic spike?"	User sim + ramp-up: pywrkr url -u 1000 -d 120 --ramp-up 10
"What's the experience like for real users?"	User sim + think time: pywrkr url -u 200 -d 300 --think-time 1.5