DIGITAL TWIN / CAPACITY INTELLIGENCE
Your next bottleneck is already in the model.
LoadTwin turns real load-test measurements into a simulation of your service — so you can challenge the assumptions behind your next deployment.
A simulation tool for understanding capacity, cost and performance risk. No account required.
Within limits. Within the example resource and latency limits.
Curiosity, meet consequence.
Change the load or the machine setup. Add traffic, add nodes, crash one. Every change is re-simulated by the real engine, on the spot.
Launch the playground when you're ready. It uses the simulation engine and the service currently saved in your browser.
A load test tells you what happened.
What about everything you haven't tested?
Production conditions rarely stand still. Workloads shift. Machines change. Dependencies slow down. LoadTwin lets you explore those scenarios using an explicit model calibrated from the evidence you already have.
Start with your measurementsThe first failure isn't always where you expect.
Within the example thresholds.
CPU breaches the 80% resource limit.
Latency alone would not tell the full story. In this documented example, resource saturation becomes the limiting factor as demand doubles.
Illustrative output reproduced from the project's documented example simulation. Not a measurement of a live production service.
THE WORK / THREE IMPORTANT DECISIONS
Complex systems.
Clearer questions.
One model. Different ways to understand how your service behaves — before you change the infrastructure behind it.
Will this system hold?
Understand if your target configuration meets latency, error and utilization limits — and which resource becomes the constraint first.
What should we change?
Compare machines, node counts and the headroom of each option before committing to the next infrastructure configuration.
What happens when things go wrong?
Explore surges, node failures and bottlenecks. See how simulated performance changes when the environment is under pressure.
Seven methods, one model, plus a live mode
Each tool asks the same calibrated twin a different question. Start anywhere; they share your service.
- DecideEvaluateDoes the target setup hold at peak?Verdict, limiting resource, capacity, breach probability
- PlanAuto-sizeWhat is the cheapest cluster that passes?Nodes per machine type and monthly cost
- StressLoad patternsDoes it survive a spike, a ramp or a full day?Time over the SLA and recovery time
- StressFailure testingWhat if a node dies or the database slows down?Before, during and after, against a healthy run
- TrustSensitivityWhich setting moves latency the most?A ranking of inputs and what to change first
- TrustValidate modelHow far can I trust the model?Predicted against measured, from k6, JMeter or Locust
- AskAsk LoadTwin“What if traffic doubles on payday?”A scenario built from your words, simulated and explained
- PlayLive modeAdd traffic, add nodes, crash one, and watch.Re-simulated on every click, on this page
THE METHOD
Make the model visible.
Make the decision defensible.
LoadTwin is built around discrete-event simulation with SimPy. It models queues and shared constraints rather than treating infrastructure capacity as a simple multiplier.
Explore model validation- 01
Calibrate
Start with observed throughput, CPU and latency from a load test.
- 02
Simulate
Change demand, resource sizing and failure conditions inside a controlled model.
- 03
Interpret
Read the predicted limits, constraints and assumptions together — not as isolated scores.
YOUR NEXT DEPLOYMENT
Know more.
Assume less.
Bring your test measurements, or begin with an example. See the model at work before making a capacity decision.