Skip to content

CodingLevel 4

Backend internals

Easy

Medium

Hard

Continue Read the guide View as list

Network & protocol parsingThe network delivers a continuous stream of bytes. You must find the message boundaries yourself.3 min Database internalsHow databases physically execute queries, like joining two tables together efficiently.3 min Storage data structuresHow modern databases achieve massive write throughput using LSM Trees.3 min Distributed algorithmsCoordinating data across multiple machines using Consistent Hashing.3 min Changing an API people already useOnce someone else's code depends on yours, the only safe way to change it is to make the new thing true before the old thing stops being true.7 min Gossip disseminationOne machine tells a few, they each tell a few, and the whole fleet knows in a handful of rounds.3 min Copy-on-write snapshotsFreeze a volume in an instant by sharing blocks, and only copy one the moment someone writes to it.3 min Anti-entropy convergenceHow distributed replicas gossip to heal silent data divergence.3 min Counting how much storage a tenant usesKeep one honest number, bytes used, and nudge it on every write and delete, so you always know whether the next write still fits.3 min Covering indexAnswering queries without ever visiting the actual table.3 min CRDT OR-set (replica merge)How two offline devices magically merge their shopping carts without conflicts.3 min Database query optimizerHow the database decides the fastest way to fetch your data.3 min Encoding and schema evolutionTurn a record into compact bytes with numbered fields, so old and new code can still read each other's data.4 min How a cost-based optimizer picks a query planThe database estimates what each possible plan would cost, then runs the cheapest one.5 min Hybrid logical clocksHow distributed databases figure out which event happened first when their wall-clocks are out of sync.3 min LSM tree compactionCleaning up the mess of append-only log files.3 min MAC addressesEvery network card is born with a permanent name tag, and a switch quietly learns who sits on which port just by reading the return address on passing mail.4 min Percent-decodingA URL can only safely carry a small alphabet, so every other byte travels disguised as %XX, decoding is just reading those disguises back into bytes.4 min Read repair across replicasA read is a chance to notice that a replica fell behind, and to quietly hand it the fresh value on the way out.3 min Read-through cacheKeep a hot copy of slow data nearby, so most reads never have to touch the database.4 min Secondary indexLooking up data when you don't know the Primary Key.3 min Sloppy quorum with hinted handoffWhen the home replicas are unreachable, accept the write on healthy backups and hand it back once the originals recover, so writes stay available during a partition.4 min Storage allocator (slab + free list)Grab one big slab of memory up front, then hand out fixed-size blocks from a free list so every alloc and free is just a pointer move.4 min Versioned storeKeep every snapshot, never the eraser, a write adds a new page instead of rubbing out the old one.3 min Reading a huge file with mmapHand the whole file to the kernel as addresses, and pay for disk reads only when you actually touch a page.4 min