Simquence

Design-Time Latency Exploration

Simulate your architecture's latency before you build it: model the work, the events and the queues, run it thousands of times and see which chain of work users would actually wait on.

Windows, macOS and Linux · free · local-first, no account

The Simquence UI showing a makespan distribution, critical-path frequencies, percentiles and the dominant critical path
The shipped checkout example, run 200 times. Most people expect the slowest back end to dominate. In most runs a 150 ms debounce holds the user up instead.

The whole workflow

01

Model

Tasks, the events that trigger them and the resources they queue behind.

02

Run

Thousands of seeded runs; queues form where concurrency runs out.

03

Read

Percentiles of how long users waited and the chain that held them up.

04

Change one thing

Raise a concurrency, remove a debounce, split a queue.

05

Compare

Same seed, same randomness: any shift is the design change.

Models without hand-written JSON

The Model Composer titled new model, showing the System pane with the default model name, schema version and entry event
Compose writes a new model, part by part.
The Model Composer titled editing checkout, showing the checkout model's System pane; its ten tasks are listed in the tree on the left
Edit changes the loaded one; export and run again on the same seed.

Questions

What is a design-time latency simulator?

A tool that predicts how long a flow takes from a model of its design rather than from running code. Simquence runs a model of tasks, events and shared resources thousands of times and reports latency percentiles from p50 to p99 with the critical path of each run.

How is it different from a profiler or APM tool?

A profiler or APM tool measures a system that already runs. Simquence works on the design before the code exists, so a structural decision can be tested while it is still cheap to change.

What can it model?

Any event-driven software: a web checkout, a desktop UI thread, a microservice fan-out, a mobile flow or an embedded pipeline. Queueing comes from each context's concurrency; debounces, backoffs and poll intervals are delays on the wiring.

Are the results reproducible?

Yes. Every run is seeded, so the same model and seed give the same answer every time. Rerun a changed model on the same seed and any difference is the design change rather than luck.

Is it free? Does it send my models anywhere?

It is free under GPL-3.0, with an LGPL-3.0 desktop interface. Models and results stay on your machine; the only outbound call is an anonymous daily check of GitHub for a newer release.

The full guide and how to read the output ship inside the app. Why it exists and who it is for.

Supporting Simquence

Free, with no paid tier and nothing held back. If it has saved you time, a contribution supports its maintenance.