ServerSloth

System design simulator

Design it. Load it. Watch it break.

Put real components (EC2, EKS, RDS, Redis, Kafka, Lambda and 206 more) on a canvas, send millions of simulated users at them, and see where your architecture fails and why.

Open the simulatorBrowse templates

Free. No sign-up. Runs in your browser.

30 ms 30 ms 30 ms Users100 × 1 req/s Busroutes · 2 ms Serverhandles · 20 ms Databasequeries · 40 ms 30 ms 30 ms 30 ms Users100 × 1 req/s Busroutes · 2 ms Serverhandles · 20 ms Databasequeries · 40 ms
RequestResponse One round trip: 6 × 30 ms network + 2 + 20 + 40 ms = 242 ms · shown 15× slower

Not a diagram. A simulation.

  1. 01

    Build

    Choose from 212+ components across AWS, Google Cloud, Azure and open source. Each one comes with realistic limits: connections, partitions and concurrency.

  2. 02

    Load

    Every request is simulated with queueing theory. Utilisation, Little’s law, cache hit ratios and consumer lag are the numbers the engine actually computes.

  3. 03

    Break

    Kill nodes, add latency, cut links or fail a zone. You get the real error message, the numbers behind it and what to try next.

Then scale it to millions

Open the simulator →
ServerSloth simulating a multi-region architecture with Route 53, load balancers, EKS and Aurora

Start from a reference architecture

All templates →
Cached web app architecture running in the simulatorStartup

Cached web app

ALB → three EC2 servers using ElastiCache (cache-aside) in front of RDS PostgreSQL. At 60 s Redis is flushed.

Event-driven microservices architecture running in the simulatorScale-up

Event-driven microservices

ALB → EKS → Kafka (MSK, 6 partitions, 60 s retention) → consumers → PostgreSQL and OpenSearch.

Multi-region architecture running in the simulatorGlobal

Multi-region

Route 53 → two zones, each with an ALB and EKS → Aurora writer in zone A with a replica in zone B. Zone A fails at 40 s.

Questions

What is ServerSloth?

ServerSloth is a free, browser-based system design simulator. You build an architecture from real components — servers, load balancers, databases, caches, queues and managed cloud services — send simulated traffic through it, and watch how it behaves and where it fails.

Is the simulation realistic?

Every component follows a documented model from queueing theory: utilisation, Little’s law, connection limits, cache hit ratios, partition limits and more. The models are simplified for learning, and each product says what it approximates.

Which cloud services can I use?

Over 200 components from AWS (EC2, EKS, RDS, Aurora, Lambda, SQS, DynamoDB…), Google Cloud, Azure, Cloudflare and open source (PostgreSQL, MySQL, Redis, Kafka, Nginx, Kubernetes…).

Is it useful for system design interviews?

Yes. Instead of only drawing boxes, you can see why a cache needs a fallback plan, why databases need connection pooling, or why Kafka parallelism is limited by partitions — with numbers you can quote.

Do I need an account?

No. ServerSloth runs entirely in your browser and saves your canvas locally.