Skilltree
Browse topics

A learning platform for engineers

Make ideas
connect.

Learn system design by connecting ideas, experimenting with real systems, and recalling what matters.

Start learning

Start with System Design. No account required.

Explore Skilltree ↓

System Design

17 concept pages

Not read Read Review due

Topics & prerequisites ↓

Branches show what to read first. Every topic is open to explore.

Topic map
17 concept pages17 interactive labs8 worked solutions

Choose your
next branch.

Start with what is here.
More ways to grow are taking shape.

Available now

System Design

From the first request to the trade-offs at scale. Read the concepts, break a simulation, and explain your decisions.

Explore System Design ↗
17 concept pages17 interactive labs8 worked solutions

More pathsNot available yet

Senior BackendIn progress
DevOps / SREComing soon
CloudComing soon
Data structures & algorithmsIn progress

Understanding
takes practice.

Read. Experiment. Recall. Review.
Make each idea earn its place.

Recall it.

Test what you remember before looking at the explanation.

Test yourself →Quiz questions include Draft content

The systems are the proof

Go on.
Break a little.

What happens when one server leaves a hash ring? Try it here. Only the keys that need a new owner move.

Open Consistent hashing ↗

The same deterministic model as the full lab. One ring point per server; 5,000 example keys. A model, not a production benchmark.

CBA3 serversSee what moves.

A key belongs to the next server clockwise. Open the full lab for virtual nodes and replicas.

Plan, executedDraft

December revenue by category · query & recording
PostgreSQL 18.6 (PGlite 0.5.9)

  1. Seq Scan on categories40 actual rows · 40 estimated per loop
  2. Hash40 actual rows · 40 estimated per loop
  3. Seq Scan on products2,000 actual rows · 2,000 estimated per loop
  4. Hash2,000 actual rows · 2,000 estimated per loop
  5. Seq Scan on orders4,001 actual rows · 4,001 estimated per loop
  6. Hash4,001 actual rows · 4,001 estimated per loop
  7. Seq Scan on order_items300,350 actual rows · 300,350 estimated per loop
  8. Hash Join12,028 actual rows · 12,017 estimated per loop
  9. Hash Join12,028 actual rows · 12,017 estimated per loop
  10. Hash Join12,028 actual rows · 12,017 estimated per loop
  11. HashAggregate40 actual rows · 40 estimated per loop
  12. Sort5 actual rows · 40 estimated per loop
  13. Limit5 actual rows · 5 estimated per loop

Read inside out: children do their work before passing rows to their parents.

Real queries. Real output.

Follow the
evidence.

This is a recorded PostgreSQL execution plan, not a picture of one. Follow each operation from table scans to the final result.

21 database pages · 12 tuning exercises Draft

Explore databases ↗

Recorded rows and estimates come from the site’s own SQL run. They describe this query and dataset, not every database.

Every connection
starts somewhere.

Browse the full map. Prerequisites are suggestions, never locked doors.

One idea.
Then the next.

Start learning
Esc
↑ ↓ to move↵ to open