System Design, Step by Step

Reference · 312 terms

Glossary

Every short form used across the series. Inside a lesson, hover or tap any dotted term to see the same entry in place.

Networking & the Internet

AnycastAnycast
The same IP address is announced from many locations; users reach the nearest one. Example: Cloudflare's 1.1.1.1 is answered by a nearby data centre wherever you are.Lesson 11 →
BGPBorder Gateway Protocol
How large networks tell each other which IP addresses they can reach. Example: The internet's map. Facebook's 2021 outage happened when its BGP routes disappeared.Lesson 1 →
CIDRClassless Inter-Domain Routing
A short way to write a range of IP addresses. Example: 10.0.0.0/16 means all addresses from 10.0.0.0 to 10.0.255.255.Lesson 1 →
DNSDomain Name System
Turns human names into IP addresses. Example: The internet's phone book: amazon.com → 52.94.236.248.Lesson 1 →
Egress / IngressEgress / Ingress
Data going out of / into a network or cloud. Example: Cloud providers usually charge for egress (data sent to users).Lesson 10 →
GeoDNSGeographic DNS
DNS returns different IPs based on where the user is. Example: Users in India get the Mumbai servers; users in Germany get Frankfurt.Lesson 11 →
IPInternet Protocol
Gives every device an address and moves packets between networks. Example: Like a house address on an envelope: 142.250.183.14.Lesson 1 →
IPv4 / IPv6IP version 4 / 6
Two address formats. IPv4 is short (and has run out); IPv6 is long, with almost unlimited addresses. Example: IPv4: 192.168.1.10; IPv6: 2001:db8::1.Lesson 1 →
ISPInternet Service Provider
The company that connects you to the internet. Example: Your broadband or mobile data provider.Lesson 1 →
L4 / L7Layer 4 / Layer 7
L4 = transport (IP + port). L7 = application (HTTP: URLs, headers). Example: An L4 load balancer sees "port 443"; an L7 one sees "GET /api/orders".Lesson 12 →
LAN / WANLocal / Wide Area Network
A LAN is a small network (home, office); a WAN connects distant locations. Example: Office Wi-Fi is a LAN; the link between two office cities is a WAN.
MTUMaximum Transmission Unit
The largest packet size a network link can carry. Example: Usually about 1,500 bytes on Ethernet. Bigger data is split into several packets.Lesson 1 →
NATNetwork Address Translation
Lets many private devices share one public IP. Example: Your home Wi-Fi router: all phones and laptops appear online as one address.Lesson 1 →
OSI modelOpen Systems Interconnection
A 7-layer model describing how network communication works. Example: "L4" and "L7" load balancers come from this model.Lesson 1 →
PoPPoint of Presence
A CDN or network location in a city. Example: A CDN with PoPs in Chennai, Delhi and Mumbai.Lesson 14 →
PortPort
A number that identifies a program on a machine. Example: 80 = HTTP, 443 = HTTPS, 22 = SSH, 5432 = PostgreSQL.Lesson 1 →
QUICoriginally "Quick UDP Internet Connections"
A modern protocol built on UDP, with reliability and encryption built in. Example: Powers HTTP/3. Pages load faster on patchy mobile networks.Lesson 2 →
RTTRound-Trip Time
Time for a message to go to a server and back. Example: India → US and back ≈ 200+ ms.Lesson 1 →
SocketSocket
One end of a network connection: IP + port + protocol. Example: Your browser tab talking to 142.250.183.14:443.Lesson 1 →
TCPTransmission Control Protocol
Reliable, ordered delivery of data. Resends anything lost. Example: A courier with tracking: every page arrives, in order. Used by websites and databases.Lesson 2 →
TTLTime To Live
How long something may be cached or kept before it expires. Example: A DNS answer with TTL 300 can be reused for 5 minutes.Lesson 1 →
UDPUser Datagram Protocol
Fast delivery with no guarantees. Lost data isn't resent. Example: Shouting across a room: fast, but some words may be missed. Used by video calls and games.Lesson 2 →
VPNVirtual Private Network
An encrypted tunnel into a private network over the internet. Example: Working from home but accessing office systems securely.

Web, HTTP & APIs

APIApplication Programming Interface
A defined way for programs to talk to each other. Example: A weather app asks a weather API: "What's the temperature in Pune?".Lesson 30 →
BFFBackend For Frontend
A small API layer tailored to one type of client. Example: A mobile BFF returns compact data; a web BFF returns detailed data.Lesson 13 →
CORSCross-Origin Resource Sharing
Lets a server allow web pages from other domains to call it. Example: Your React app on app.com calling api.com.Lesson 3 →
CRUDCreate, Read, Update, Delete
The four basic operations on data. Example: POST, GET, PUT/PATCH, DELETE in REST.Lesson 30 →
ETagEntity Tag
A fingerprint of a resource's version. Example: "Has this changed since version abc123?" If not → 304 Not Modified.Lesson 3 →
GraphQLGraph Query Language
Lets clients ask for exactly the fields they need in one request. Example: A mobile app fetches name + last 3 orders + thumbnails in one call.Lesson 32 →
gRPC(Google's) RPC framework
Fast, typed RPC over HTTP/2 using Protocol Buffers. Example: Internal service-to-service calls in microservices.Lesson 31 →
HSTSHTTP Strict Transport Security
Tells browsers to always use HTTPS for a site. Example: Even if you type http://bank.com, the browser goes to https://.Lesson 3 →
HTTPHyperText Transfer Protocol
The request/response language of the web. Example: Browser: GET /products/991 → server: 200 OK + data.Lesson 3 →
HTTP/1.1, HTTP/2, HTTP/3HTTP/1.1, HTTP/2, HTTP/3
HTTP versions: one request at a time per connection → many at once (multiplexing) → over QUIC. Example: HTTP/2 loads all of a page's images over one connection.Lesson 3 →
HTTPSHTTP Secure
HTTP encrypted with TLS. Example: The padlock in your browser's address bar.Lesson 3 →
IdempotencyIdempotency
Doing something twice has the same effect as doing it once. Example: A lift button: pressing it 5 times still calls one lift. Retrying a payment with the same key charges once.Lesson 34 →
JSONJavaScript Object Notation
A simple text format for data. Example: {"name": "Asha", "age": 28}.Lesson 30 →
MQTTMessage Queuing Telemetry Transport
A lightweight publish/subscribe protocol for devices. Example: Smart home sensors sending temperature readings.Lesson 33 →
OpenAPIformerly Swagger
A standard file describing a REST API. Example: Generates docs and client SDKs automatically.Lesson 30 →
Pagination / CursorPagination / Cursor
Returning results a page at a time; a cursor is a bookmark to the next page. Example: "Load more" on a feed.Lesson 34 →
ProtobufProtocol Buffers
A compact binary data format with a schema. Example: The same data as JSON, but much smaller and faster to parse.Lesson 31 →
Rate limitRate limit
A cap on how many requests a client can make in a time window. Example: "100 requests per minute on the Free plan".Lesson 43 →
RESTRepresentational State Transfer
An API style built on resources (nouns) and HTTP methods (verbs). Example: GET /users/42, POST /orders, DELETE /orders/7.Lesson 30 →
RPCRemote Procedure Call
Calling a function on another machine as if it were local. Example: inventory.CheckStock("991") runs on another server.Lesson 31 →
SDKSoftware Development Kit
A library that makes an API easy to use in a language. Example: Stripe's Python SDK for taking payments.
SSEServer-Sent Events
A one-way stream of updates from server to browser over HTTP. Example: AI answers appearing word by word; live scores.Lesson 33 →
Status codesStatus codes
Three-digit results of an HTTP request. Example: 200 OK, 201 Created, 404 Not Found, 429 Too Many Requests, 503 Unavailable.Lesson 3 →
URL / URIUniform Resource Locator / Identifier
The address of a resource on the web. Example: https://shop.com/products/991?color=blue.Lesson 1 →
WebhookWebhook
A server calls your URL when something happens. Example: Payment provider calls /webhooks/payment when a payment succeeds.Lesson 34 →
WebRTCWeb Real-Time Communication
Peer-to-peer audio, video and data in browsers. Example: Video calls in the browser.Lesson 33 →
WebSocketWebSocket
A persistent, two-way connection between browser and server. Example: Chat messages arriving instantly.Lesson 33 →

Operating Systems & Concurrency

C10K10,000 concurrent connections
The classic challenge of handling many simultaneous connections. Example: Solved by event loops and non-blocking I/O.Lesson 4 →
Concurrency / ParallelismConcurrency / Parallelism
Juggling many tasks / doing many tasks at the same instant. Example: One cook switching dishes vs several cooks at once.Lesson 4 →
Connection poolConnection pool
A reusable set of open connections. Example: 20 database connections shared by all requests.Lesson 5 →
CPUCentral Processing Unit
Runs program instructions. Example: More cores = more work in parallel.Lesson 4 →
DeadlockDeadlock
Two tasks each wait for the other forever. Example: Two cars blocking each other on a narrow bridge.Lesson 4 →
Event loopEvent loop
One thread handling many tasks by never waiting idly. Example: Node.js and NGINX serving thousands of connections.Lesson 4 →
GILGlobal Interpreter Lock
Lets only one thread run Python code at a time (in standard CPython). Example: Python uses multiple processes for CPU-heavy work.Lesson 4 →
I/OInput/Output
Reading/writing disks, network, files. Example: Waiting for a database reply is I/O.Lesson 4 →
Mutex / LockMutual exclusion
Lets only one thread use something at a time. Example: A single key to the storeroom.Lesson 4 →
ProcessProcess
A running program with its own memory. Example: Each Chrome tab is roughly its own process.Lesson 4 →
Race conditionRace condition
A bug where the result depends on the timing of concurrent actions. Example: Two people booking the last seat at the same moment.Lesson 4 →
RAMRandom Access Memory
Fast, temporary working memory. Example: Redis keeps its data in RAM, which is why it's fast.Lesson 10 →
SSD / HDDSolid State Drive / Hard Disk Drive
Permanent storage. SSD has no moving parts and is much faster. Example: Databases run far better on SSDs.Lesson 10 →
ThreadThread
A worker inside a process that shares its memory. Example: A web server handling 200 requests with 200 threads.Lesson 4 →

Performance, Capacity & Reliability Numbers

Availability / "nines"Availability / "nines"
% of time (or requests) the system works. Example: 99.9% ≈ 8.8 hours of downtime per year.Lesson 9 →
BandwidthBandwidth
Maximum data capacity of a link. Example: A 1 Gbps network connection.Lesson 8 →
Burn rateBurn rate
How fast you're using the error budget. Example: Burn rate 14.4 = whole monthly budget gone in ~2 days.Lesson 47 →
DAU / MAUDaily / Monthly Active Users
Number of users active per day / month. Example: "10M DAU" is the starting point of most estimates.Lesson 10 →
Error budgetError budget
Allowed failure = 100% − SLO. Example: 99.9% SLO on 10M requests → 10,000 failures allowed.Lesson 47 →
HeadroomHeadroom
Spare capacity for spikes and failures. Example: Running servers at 60% CPU, not 95%.Lesson 8 →
LatencyLatency
Time to complete one request. Example: "The API responds in 80 ms.".Lesson 8 →
Little's LawLittle's Law
Items in the system = arrival rate × time each spends. Example: 500 req/s × 0.04 s = 20 requests in flight.Lesson 8 →
MTBFMean Time Between Failures
Average time between failures. Example: Failing once a month = MTBF ≈ 720 hours.Lesson 9 →
MTTRMean Time To Recovery (or Repair)
Average time to recover from a failure. Example: Rollback in 5 minutes instead of 4 hours.Lesson 9 →
N+1 (capacity)N+1 (capacity)
One more server than needed, to survive a failure. Example: Need 4 servers? Run 5.Lesson 40 →
p50 / p95 / p9950th / 95th / 99th percentile
The latency that 50% / 95% / 99% of requests are faster than. Example: p99 = 800 ms means 1 in 100 requests takes longer than 800 ms.Lesson 8 →
QPS / RPS / TPSQueries / Requests / Transactions Per Second
Measures of load. Example: "Peak 15,000 QPS on the feed.".Lesson 10 →
RPORecovery Point Objective
How much data you can afford to lose. Example: RPO 5 minutes = backups/logs at least every 5 minutes.Lesson 45 →
RTORecovery Time Objective
How long you can afford to be down. Example: RTO 1 hour = must be back online within an hour.Lesson 45 →
SLAService Level Agreement
A promise to customers with penalties. Example: "Below 99.9% monthly uptime, you get a 10% credit.".Lesson 47 →
SLIService Level Indicator
What you measure about user experience. Example: % of checkout requests that succeed.Lesson 47 →
SLOService Level Objective
Your target for an SLI. Example: 99.9% of checkouts succeed over 28 days.Lesson 47 →
ThroughputThroughput
Work done per unit of time. Example: "The API handles 5,000 requests per second.".Lesson 8 →

Databases & Data

ACIDAtomicity, Consistency, Isolation, Durability
Guarantees that transactions are reliable. Example: A bank transfer: both debit and credit happen, or neither.Lesson 21 →
B-treeBalanced tree
A sorted, shallow tree used by most database indexes. Example: Finds 1 row among 1 billion in ~3–4 steps.Lesson 22 →
BASEBasically Available, Soft state, Eventually consistent
A looser model favouring availability. Example: Many NoSQL systems.Lesson 21 →
Bloom filterBloom filter
A tiny structure that says "definitely not here" or "maybe here". Example: Skips reading files that can't contain a key.Lesson 16 →
CDCChange Data Capture
Streams every database change as an event. Example: Debezium sends DB changes to Kafka → search index.Lesson 19 →
Data warehouseData warehouse
A database built for analytics. Example: BigQuery, Snowflake, Redshift, ClickHouse.Lesson 29 →
DenormalizationDenormalization
Copying data on purpose for faster reads. Example: Storing author_name in each post.Lesson 23 →
ETL / ELTExtract, Transform, Load
Moving data from sources into a warehouse. Example: Nightly copy of orders into BigQuery.
IndexIndex
A structure that makes lookups fast. Example: The index at the back of a textbook.Lesson 5 →
Isolation levelsIsolation levels
How much concurrent transactions can see of each other. Example: Read Committed, Repeatable Read, Serializable.Lesson 21 →
LSM-treeLog-Structured Merge-tree
A write-optimised storage structure (memtable + sorted files + compaction). Example: Used by Cassandra, RocksDB, ScyllaDB.Lesson 22 →
Materialized viewMaterialized view
A stored, precomputed query result. Example: A daily sales summary table.Lesson 23 →
MVCCMulti-Version Concurrency Control
Keeps several versions of rows so readers don't block writers. Example: PostgreSQL shows each transaction a consistent snapshot.Lesson 21 →
N+1 problemN+1 problem
One query plus N more, one per item. Example: Loading 20 posts, then 20 separate author queries.Lesson 5 →
NormalizationNormalization
Storing each fact once, to avoid inconsistencies. Example: Customer city stored only in customers.Lesson 23 →
NoSQL"Not only SQL"
Non-relational databases: key-value, document, wide-column, graph. Example: Redis, MongoDB, Cassandra, Neo4j.Lesson 20 →
OLAPOnline Analytical Processing
Big queries over lots of data. Example: "Revenue by city for last year.".Lesson 5 →
OLTPOnline Transaction Processing
Many small, fast reads/writes. Example: Placing orders, updating profiles.Lesson 5 →
ORMObject-Relational Mapper
Maps database rows to code objects. Example: Prisma, Django ORM, Hibernate.Lesson 5 →
PITRPoint-In-Time Recovery
Restore a database to any exact moment. Example: Restore to 10:41:59, one second before a bad DELETE.Lesson 45 →
PK / FKPrimary Key / Foreign Key
PK uniquely identifies a row; FK points to a row in another table. Example: orders.user_id (FK) → users.id (PK).Lesson 5 →
RDBMSRelational Database Management System
A database of tables linked by keys. Example: PostgreSQL, MySQL, SQL Server.Lesson 5 →
SchemaSchema
The structure of data (tables, fields, types). Example: users(id, email, name).Lesson 20 →
Snowflake IDSnowflake ID
64-bit ID = timestamp + machine ID + sequence. Example: Unique, time-ordered IDs generated on any server.Lesson 25 →
SQLStructured Query Language
The language for relational databases. Example: SELECT name FROM users WHERE id = 42;.Lesson 5 →
SSTableSorted String Table
An immutable, sorted file in LSM storage. Example: Written when the in-memory table fills up.Lesson 22 →
TransactionTransaction
A group of changes that succeed or fail together. Example: BEGIN … COMMIT.Lesson 21 →
UUIDUniversally Unique Identifier
A 128-bit ID that's unique without coordination. Example: 7f3c9a2e-4b1d-4e8a-9f0c-2d6b1e5a8c33.Lesson 5 →
WALWrite-Ahead Log
Changes are written to a log before the data files. Example: Makes crash recovery and replication possible.Lesson 21 →

Distributed Systems

2PCTwo-Phase Commit
Coordinates a transaction across systems (prepare, then commit). Example: Slow and fragile; often avoided in microservices.Lesson 21 →
CAP theoremConsistency, Availability, Partition tolerance
During a network partition, choose consistency or availability. Example: Bank: refuse (C) vs allow a possibly wrong balance (A).Lesson 27 →
ConsensusConsensus
Getting machines to agree on one value despite failures. Example: Electing exactly one leader.Lesson 24 →
Consistent hashingConsistent hashing
Places keys and servers on a ring so few keys move when servers change. Example: Adding a cache server moves only ~1/N of keys.Lesson 26 →
CRDTConflict-free Replicated Data Type
Data types that always merge correctly. Example: Collaborative counters or shared documents.Lesson 28 →
Eventual consistencyEventual consistency
Copies agree eventually, if updates stop. Example: Like counts that differ briefly between devices.Lesson 28 →
FailoverFailover
Switching to a backup when the main one fails. Example: A replica becomes the new primary.Lesson 24 →
FencingFencing
Stops an old leader from writing after failover. Example: Revoking the old primary's access.Lesson 24 →
GossipGossip
Nodes share information with random peers, spreading it like rumours. Example: Cassandra nodes learning cluster membership.
HLCHybrid Logical Clock
Combines wall-clock time and a counter for ordering. Example: Used by CockroachDB.Lesson 28 →
Hot key / hot spotHot key / hot spot
One key or shard getting far more traffic than others. Example: A celebrity's profile.Lesson 16 →
Leader / Followeralso Primary / Replica
The leader accepts writes; followers copy them. Example: All writes go to the primary; reads can go to replicas.Lesson 24 →
LinearizabilityLinearizability
The strongest consistency: acts like one copy. Example: Once anyone sees the new value, nobody sees the old.Lesson 28 →
LWWLast-Write-Wins
Keep the update with the latest timestamp. Example: Simple conflict resolution; may lose the other write.Lesson 28 →
PACELCPartition → Availability or Consistency; Else → Latency or Consistency
Adds the everyday latency vs consistency trade-off. Example: Strong consistency across continents is slower.Lesson 27 →
QuorumQuorum
A majority (or W + R > N) must agree. Example: 3 replicas: write to 2, read from 2 → always see the latest.Lesson 24 →
Raft / PaxosRaft / Paxos
Well-known consensus algorithms. Example: Used by etcd, Consul, CockroachDB.Lesson 24 →
Read-your-writesRead-your-writes
You always see your own changes. Example: Your new comment doesn't disappear on refresh.Lesson 28 →
ReplicationReplication
Keeping copies of data on several machines. Example: A primary database plus 3 read replicas.Lesson 24 →
Replication lagReplication lag
How far behind a replica is. Example: You update your bio, refresh, and see the old one.Lesson 24 →
SagaSaga
A sequence of local transactions with compensating actions. Example: Payment fails → release stock → cancel order.Lesson 39 →
Shard keyShard key
The field that decides which shard holds a row. Example: tenant_id in a SaaS app.Lesson 25 →
Sharding / PartitioningSharding / Partitioning
Splitting data across machines. Example: Users 1–1M on shard 1, 1M–2M on shard 2.Lesson 25 →
Split brainSplit brain
Two nodes both think they're the leader. Example: Both accept writes, so the data diverges.Lesson 24 →
Vector clockVector clock
Tracks versions per node to detect concurrent updates. Example: Amazon's Dynamo used them for cart conflicts.Lesson 28 →
Virtual nodesvnodes
Each server appears many times on the hash ring. Example: Spreads load evenly.Lesson 26 →

Caching & Storage

BlobBinary Large Object
Any large binary file. Example: A video or image.Lesson 17 →
Block storageBlock storage
A raw disk attached to a server. Example: AWS EBS volume for a database.Lesson 17 →
BM25Best Match 25
A standard formula for ranking search results. Example: Default ranking in Elasticsearch.Lesson 19 →
CacheCache
A fast copy of frequently used data. Example: Salt kept on the kitchen counter instead of in the storeroom.Lesson 15 →
Cache hit / missCache hit / miss
Found in cache / not found. Example: Hit: 1 ms from Redis. Miss: 50 ms from the database.Lesson 15 →
Cache stampedethundering herd
Many requests rebuild the same expired key at once. Example: A popular key expires → 5,000 DB queries in 1 second.Lesson 16 →
Cache-asidelazy loading
App reads cache → on miss, loads DB and fills cache. Example: The most common caching pattern.Lesson 16 →
CDNContent Delivery Network
Servers worldwide that cache content near users. Example: Images load from a server in your city.Lesson 14 →
Erasure codingErasure coding
Splits data into pieces + parity so it survives lost pieces. Example: Object storage durability with less space than 3 full copies.Lesson 17 →
Hit ratioHit ratio
% of requests served by the cache. Example: 95% hit ratio → only 5% reach the database.Lesson 15 →
Inverted indexInverted index
Maps words to the documents containing them. Example: "shoe" → docs 1, 3. Powers search engines.Lesson 19 →
Key-value storeKV store
Stores values by key. Example: session:abc → {user: 42}.Lesson 20 →
LRU / LFULeast Recently / Frequently Used
Eviction rules for full caches. Example: LRU removes the item not used for the longest time.Lesson 16 →
MemcachedMemcached
A simple, fast in-memory cache. Example: Facebook's huge caching layer.Lesson 15 →
Object storageObject storage
Stores files ("objects") by key over HTTP, at huge scale. Example: Amazon S3, Google Cloud Storage.Lesson 17 →
Pre-signed URLPre-signed URL
A temporary, signed link to upload/download one file. Example: Browser uploads a photo directly to S3.Lesson 17 →
RedisRemote Dictionary Server
An in-memory data store: cache, counters, sessions, queues. Example: Storing sessions and rate-limit counters.Lesson 15 →
S3Simple Storage Service
Amazon's object storage (and the de facto API standard). Example: Profile photos and backups.Lesson 17 →
Write-through / Write-behindWrite-through / Write-behind
Write to cache and DB together / write to cache first, DB later. Example: Write-behind for view counters.Lesson 16 →

Messaging & Events

AckAcknowledgement
"I received / processed this." Example: Worker acks after sending the email.Lesson 18 →
At-most-once / At-least-once / Exactly-onceAt-most-once / At-least-once / Exactly-once
Delivery guarantees: may lose / may duplicate / effectively once. Example: Payments need at-least-once + dedup.Lesson 37 →
BackpressureBackpressure
Slowing producers when consumers can't keep up. Example: Rejecting non-urgent jobs when the queue is too long.Lesson 18 →
Competing consumersCompeting consumers
Many workers share one queue; each message goes to one worker. Example: 5 email workers processing one queue.Lesson 18 →
Consumer groupConsumer group
Consumers sharing the work of reading a topic. Example: Group "billing" and group "analytics" each read every event.Lesson 36 →
Consumer lagConsumer lag
How far behind a consumer is. Example: Rising lag = consumers can't keep up.Lesson 36 →
CQRSCommand Query Responsibility Segregation
Separate models for writes and reads. Example: Writes to PostgreSQL; reads from a search index.Lesson 39 →
Cronfrom "chronos", time
Runs jobs on a schedule. Example: "Every night at 2 AM, clean expired sessions.".Lesson 38 →
DLQDead-Letter Queue
Where messages go after failing too many times. Example: Malformed messages parked for engineers to inspect.Lesson 38 →
EOSExactly-Once Semantics
Kafka's feature for exactly-once processing within Kafka. Example: Read → process → write back to Kafka atomically.Lesson 37 →
Event sourcingEvent sourcing
Store every change as an event; state = replay of events. Example: A bank ledger.Lesson 39 →
Fan-outFan-out
One event/message spread to many recipients. Example: One post → written into 200 followers' feeds.Lesson 64 →
KafkaApache Kafka
A distributed, durable log for event streaming. Example: LinkedIn's activity events; CDC pipelines.Lesson 36 →
MQMessage Queue
Holds messages until workers process them. Example: Order placed → email job in the queue.Lesson 18 →
OffsetOffset
A message's position in a partition. Example: "I've processed up to offset 1,041.".Lesson 36 →
Outbox patterntransactional outbox
Save data + "event to publish" in one DB transaction. Example: Solves "DB updated but event not sent".Lesson 37 →
Producer / ConsumerProducer / Consumer
Sends messages / processes messages. Example: API = producer; email worker = consumer.Lesson 18 →
Pub/SubPublish/Subscribe
One message delivered to every subscriber. Example: "Order placed" → email, warehouse and analytics all get it.Lesson 35 →
Topic / PartitionTopic / Partition
A named stream / an ordered slice of it. Example: Topic orders, partitioned by order ID.Lesson 36 →

Reliability & Resilience Patterns

3-2-1 rule3-2-1 rule
3 copies, 2 media types, 1 off-site. Example: Database + backup in same cloud + backup in another region.Lesson 45 →
Active-active / Active-passiveActive-active / Active-passive
Both copies serve traffic / one waits as backup. Example: Two regions serving users vs a standby region.Lesson 40 →
Blast radiusBlast radius
How much is affected by one failure. Example: A bad deploy in one "cell" affects only that cell's users.Lesson 40 →
BulkheadBulkhead
Separate resource pools so one failure can't sink everything. Example: A ship's watertight compartments.Lesson 42 →
Chaos engineeringChaos engineering
Deliberately injecting failures to test resilience. Example: Netflix's Chaos Monkey kills random servers.Lesson 40 →
Circuit breakerCircuit breaker
Stops calling a failing service for a while. Example: Like a home electrical breaker tripping.Lesson 42 →
DRDisaster Recovery
Plans and systems to recover from major disasters. Example: Restoring in another region after a data-centre fire.Lesson 45 →
Exponential backoffExponential backoff
Waiting longer after each failure. Example: 100 ms, 200 ms, 400 ms….Lesson 41 →
FallbackFallback
What to return when something fails. Example: Show "popular items" when recommendations are down.Lesson 42 →
Graceful degradationGraceful degradation
Turning off non-essential features under stress. Example: Hide reviews, keep checkout working.Lesson 44 →
HAHigh Availability
Designing to minimise downtime. Example: Servers in 3 zones with automatic failover.Lesson 40 →
Health checkHealth check
Regularly checking whether an instance works. Example: Load balancer calls /health every 5 seconds.Lesson 12 →
JitterJitter
Random extra delay so clients don't retry together. Example: Waits spread between 0 and 400 ms.Lesson 41 →
Kill switch / Feature flagKill switch / Feature flag
A switch to turn a feature on/off without deploying. Example: Disable a broken feature instantly.Lesson 44 →
Load sheddingLoad shedding
Rejecting excess requests to protect the system. Example: Returning 503 to low-priority traffic during overload.Lesson 44 →
RedundancyRedundancy
Having spares of critical parts. Example: Two load balancers, three app servers.Lesson 40 →
RetryRetry
Trying a failed operation again. Example: Retry once after a network blip.Lesson 41 →
SPOFSingle Point Of Failure
A part whose failure brings everything down. Example: One database with no replica.Lesson 40 →
Static stabilityStatic stability
Keeps working during a failure without needing changes. Example: Enough servers already running in each zone.Lesson 40 →
TimeoutTimeout
Maximum time to wait for a response. Example: Give up on a call after 500 ms.Lesson 41 →
Token bucket / Leaky bucketToken bucket / Leaky bucket
Rate-limiting algorithms. Example: 20-token bucket refilled at 10/s allows bursts.Lesson 43 →

Observability & Operations

APMApplication Performance Monitoring
Tools for tracing and app performance. Example: Datadog, New Relic, Dynatrace.Lesson 46 →
CardinalityCardinality
Number of unique label combinations in metrics. Example: Adding user_id as a label creates millions of series.Lesson 46 →
Golden SignalsGolden Signals
Latency, traffic, errors, saturation. Example: The four core metrics for any service.Lesson 46 →
ICIncident Commander
The person coordinating an incident. Example: Makes decisions; doesn't debug.Lesson 48 →
Incident / SEVSeverity level
An unplanned problem, ranked by impact. Example: SEV1 = payments down for many users.Lesson 48 →
LogsLogs
Timestamped records of events. Example: {"level":"error","msg":"payment failed","order":"5521"}.Lesson 46 →
MetricsMetrics
Numbers measured over time. Example: Requests per second, p99 latency, error rate.Lesson 46 →
ObservabilityObservability
Understanding a system's internal state from its outputs. Example: Answering "why is checkout slow for Android users?".Lesson 46 →
On-callOn-call
Engineers responsible for responding to alerts. Example: Weekly rotation with primary and secondary.Lesson 48 →
OTelOpenTelemetry
Open standard for producing logs, metrics and traces. Example: Instrument once, send data to any tool.Lesson 46 →
Postmortemincident review
A written analysis of what happened and how to prevent it. Example: Blameless report with action items.Lesson 48 →
REDRate, Errors, Duration
Key metrics for request-driven services. Example: 500 req/s, 0.2% errors, p99 180 ms.Lesson 46 →
RUMReal User Monitoring
Measures what real users' browsers/apps experience. Example: Page load times by country and device.Lesson 48 →
RunbookRunbook
Step-by-step instructions for handling an alert. Example: "If checkout errors spike: check deploys, then…".Lesson 48 →
SRESite Reliability Engineering
Applying software engineering to operations. Example: Google's SLOs, error budgets, postmortems.Lesson 47 →
Synthetic monitoringSynthetic monitoring
Scripted "robot users" that test the site constantly. Example: A bot that logs in and checks out every minute.Lesson 48 →
Trace ID / Correlation IDTrace ID / Correlation ID
An ID shared across all logs and spans of one request. Example: Search logs by trace_id to follow one checkout.Lesson 46 →
Traces / SpansTraces / Spans
The path of one request across services / one step in it. Example: See that the payment call took 3 of the 3.2 seconds.Lesson 46 →
USEUtilisation, Saturation, Errors
Key metrics for resources (CPU, disk, network). Example: CPU 70%, 12 requests waiting, 0 errors.Lesson 46 →

Security & Identity

ACLAccess Control List
A list of who can do what on a resource. Example: File shared with Asha (edit) and Ravi (view).Lesson 49 →
AESAdvanced Encryption Standard
Fast symmetric encryption. Example: Encrypting files and database fields.Lesson 51 →
Argon2 / bcryptArgon2 / bcrypt
Slow, salted password-hashing algorithms. Example: Storing passwords safely.Lesson 49 →
AuthNAuthentication
Proves who you are. Example: Logging in with password + OTP.Lesson 49 →
AuthZAuthorization
Decides what you're allowed to do. Example: You can see your invoices, not other people's.Lesson 49 →
CookieCookie
Small data the browser sends with each request to a site. Example: sid=k9Fq…; HttpOnly; Secure; SameSite=Lax.Lesson 49 →
CSRFCross-Site Request Forgery
Tricks your browser into sending requests with your cookies. Example: A hidden form that transfers money.Lesson 49 →
CVECommon Vulnerabilities and Exposures
A public ID for a known vulnerability. Example: Log4Shell = CVE-2021-44228.Lesson 52 →
DDoSDistributed Denial of Service
Flooding a service with traffic from many machines. Example: The 2016 Dyn attack.Lesson 1 →
E2EEEnd-to-End Encryption
Only sender and recipient can read messages. Example: WhatsApp chats.Lesson 51 →
Envelope encryptionEnvelope encryption
Data encrypted with a data key; data key encrypted with a master key. Example: Cheap key rotation.Lesson 51 →
GDPR / PCI DSSGeneral Data Protection Regulation / Payment Card Industry Data Security Standard
Privacy law (EU) / card security standard. Example: Rules on consent, deletion and card data.Lesson 51 →
HMACHash-based Message Authentication Code
Proves a message wasn't changed and came from a key holder. Example: Signing webhooks.Lesson 34 →
IAMIdentity and Access Management
Managing users, roles and permissions (especially in cloud). Example: An AWS role allowing a service to read one bucket.Lesson 51 →
IDOR / BOLAInsecure Direct Object Reference / Broken Object-Level Authorization
Accessing other users' objects by changing an ID. Example: /invoices/1002 shows someone else's invoice.Lesson 49 →
JWTJSON Web Token
A signed token containing claims about a user. Example: header.payload.signature sent in Authorization: Bearer.Lesson 49 →
KMS / HSMKey Management Service / Hardware Security Module
Securely store and use encryption keys. Example: AWS KMS encrypts data keys; keys never leave it.Lesson 51 →
Least privilegeLeast privilege
Give only the minimum permissions needed. Example: The app's DB user can't drop tables.Lesson 49 →
MFA / 2FAMulti-Factor / Two-Factor Authentication
Requires more than a password. Example: Password + code from an app.Lesson 49 →
mTLSMutual TLS
Both sides present certificates. Example: Service A and service B verify each other.Lesson 51 →
OAuth 2.0Open Authorization
Lets apps get limited access to your data without your password. Example: "Allow this app to read your Google Photos".Lesson 50 →
OIDCOpenID Connect
Adds login (identity) on top of OAuth 2.0. Example: "Sign in with Google".Lesson 50 →
OTP / TOTPOne-Time Password / Time-based OTP
A short-lived code. Example: 6-digit SMS code; authenticator app code that changes every 30 s.Lesson 49 →
OWASPOpen Worldwide Application Security Project
Community known for the Top 10 web security risks. Example: Broken access control, injection, SSRF….Lesson 52 →
PasskeyWebAuthn / FIDO2
Passwordless login using device cryptography. Example: Unlock with fingerprint; nothing to phish.Lesson 49 →
PIIPersonally Identifiable Information
Data that identifies a person. Example: Name, phone, email, address, ID numbers.Lesson 46 →
PKCEProof Key for Code Exchange
Protects the OAuth code exchange from interception. Example: Required for mobile and single-page apps.Lesson 50 →
PKI / CAPublic Key Infrastructure / Certificate Authority
The system of trusted certificates / an organisation that issues them. Example: Let's Encrypt issues free certificates.Lesson 3 →
RBAC / ABAC / ReBACRole- / Attribute- / Relationship-Based Access Control
Ways to model permissions. Example: RBAC: "editors can edit". ReBAC: "team members can view team docs".Lesson 49 →
RSA / ECCRivest–Shamir–Adleman / Elliptic Curve Cryptography
Public-key cryptography. Example: Key exchange and signatures in TLS.Lesson 51 →
SAMLSecurity Assertion Markup Language
An older XML-based SSO standard, common in enterprises. Example: Logging into company tools via Okta.Lesson 50 →
SAST / DASTStatic / Dynamic Application Security Testing
Scanning code / testing the running app for vulnerabilities. Example: CI scans every pull request.Lesson 52 →
SBOMSoftware Bill of Materials
A list of all components in your software. Example: Quickly finding every service that uses Log4j.Lesson 52 →
SessionSession
Server-side record of a logged-in user. Example: Session ID in a cookie → user data in Redis.Lesson 49 →
SHA-256Secure Hash Algorithm
A one-way fingerprint of data. Example: Checking a download wasn't altered.Lesson 51 →
SQLiSQL Injection
Input is treated as SQL code. Example: ' OR '1'='1 returns every user.Lesson 52 →
SSLSecure Sockets Layer
The old, insecure predecessor of TLS (the name is still used loosely). Example: "SSL certificate" usually means a TLS certificate.Lesson 3 →
SSOSingle Sign-On
Log in once, access many apps. Example: Company login gives access to email, Slack and HR tools.Lesson 50 →
SSRFServer-Side Request Forgery
Tricks a server into requesting internal resources. Example: Fetching cloud metadata credentials.Lesson 52 →
TDETransparent Data Encryption
Database encrypts its files and backups automatically. Example: Protects stolen disks, but not someone who can query the DB.Lesson 51 →
TLSTransport Layer Security
Encrypts data between two parties and verifies the server. Example: The "S" in HTTPS.Lesson 3 →
TokenisationTokenisation
Replaces sensitive data with a meaningless token. Example: Saved card = tok_9f3a…, not the real number.Lesson 51 →
WAFWeb Application Firewall
Filters malicious HTTP requests. Example: Blocking common SQL injection patterns at the edge.Lesson 14 →
XSSCross-Site Scripting
Attacker's script runs in other users' browsers. Example: A malicious comment that steals tokens.Lesson 52 →
Zero trustZero trust
Never trust the network; verify every connection. Example: Encrypt and authenticate even internal traffic.Lesson 51 →

Architecture, Containers & Deployment

ADRArchitecture Decision Record
A short document recording a design decision. Example: "We chose PostgreSQL because…".Lesson 6 →
API gatewayAPI gateway
Front door for APIs: auth, rate limits, routing. Example: /orders → order service, /users → user service.Lesson 13 →
Blue-green deploymentBlue-green deployment
Two environments; switch traffic from old to new. Example: Instant rollback by switching back.Lesson 58 →
Canary releaseCanary release
Send a small % of traffic to the new version first. Example: 5% → 25% → 100% if metrics look good.Lesson 58 →
CDContinuous Delivery / Deployment
Always ready to deploy / automatically deploy. Example: Passing changes roll out to production.Lesson 58 →
CIContinuous Integration
Automatically build and test every change. Example: Every pull request runs tests.Lesson 58 →
ContainerContainer
A lightweight, isolated package of an app and its dependencies. Example: The same image runs on a laptop and in production.Lesson 55 →
DDDDomain-Driven Design
Designing software around business domains. Example: "Orders" and "Payments" as separate bounded contexts.Lesson 53 →
Deployment / StatefulSet / DaemonSetDeployment / StatefulSet / DaemonSet
Kubernetes workload types: stateless apps / stateful apps / one per node. Example: Deployment for an API; StatefulSet for a database.Lesson 56 →
DockerDocker
The popular tool for building and running containers. Example: docker build, docker run.Lesson 55 →
DORA metricsDevOps Research and Assessment
Deployment frequency, lead time, change failure rate, time to restore. Example: Measuring delivery performance.Lesson 58 →
GitOpsGitOps
Git is the source of truth; tools sync the system to it. Example: Argo CD deploys whatever is merged.Lesson 58 →
HelmHelm
A package manager for Kubernetes applications. Example: Install Redis with one Helm chart.Lesson 56 →
HPAHorizontal Pod Autoscaler
Adds/removes pods based on load. Example: Scale from 3 to 30 pods when CPU passes 65%.Lesson 56 →
IaCInfrastructure as Code
Define infrastructure in code files. Example: Terraform creates the network, database and buckets.Lesson 57 →
Image / RegistryImage / Registry
Container template / place where images are stored. Example: myshop/api:1.4.2 pushed to Docker Hub or ECR.Lesson 55 →
IngressIngress
Routes external HTTP traffic into a Kubernetes cluster. Example: shop.com/api → orders service.Lesson 56 →
K8sKubernetes ("K" + 8 letters + "s")
Orchestrates containers across many machines. Example: Keeps 5 replicas running and restarts crashed ones.Lesson 56 →
LBLoad Balancer
Spreads traffic across servers. Example: 3 app servers behind one address.Lesson 12 →
MicroservicesMicroservices
Many small, independently deployable services. Example: Separate user, order and payment services.Lesson 53 →
Modular monolithModular monolith
One deployable with strong internal boundaries. Example: Shopify's large Rails app.Lesson 53 →
MonolithMonolith
One application deployed as a single unit. Example: One Rails or Django app with one database.Lesson 53 →
OCIOpen Container Initiative
Standards for container images and runtimes. Example: Images built with Docker run on Kubernetes.Lesson 55 →
PodPod
The smallest deployable unit in Kubernetes. Example: One app container (plus sidecars).Lesson 56 →
Reverse proxyReverse proxy
Sits in front of servers: TLS, caching, compression, routing. Example: NGINX in front of a Node.js app.Lesson 13 →
Rolling updateRolling update
Replace instances a few at a time. Example: Kubernetes' default deploy strategy.Lesson 58 →
Service discoveryService discovery
How services find each other's current addresses. Example: Calling http://payment-service instead of an IP.Lesson 54 →
Service meshService mesh
Sidecar proxies that handle traffic between services. Example: Istio or Linkerd providing mTLS, retries and metrics.Lesson 13 →
SidecarSidecar
A helper container running next to the main one. Example: An Envoy proxy beside each service.Lesson 54 →
SOAService-Oriented Architecture
An earlier style of building systems from services. Example: Enterprise systems with a shared service bus.
Strangler figStrangler fig
Gradually replacing a legacy system piece by piece. Example: Moving payments out of a monolith first.Lesson 53 →
VMVirtual Machine
A full virtual computer with its own OS. Example: An EC2 instance.Lesson 55 →

Cloud

AZAvailability Zone
An independent data centre (or group) within a region. Example: Running servers in 3 AZs survives one data-centre failure.Lesson 57 →
Cold startCold start
Extra delay when a serverless function starts fresh. Example: First request after idle takes 1 second.Lesson 57 →
EC2 / EKS / ECSElastic Compute Cloud / Elastic Kubernetes Service / Elastic Container Service
AWS virtual machines / managed Kubernetes / managed containers. Example: Common AWS compute options.Lesson 57 →
Edge locationEdge location
Small sites near users for CDN and DNS. Example: Hundreds worldwide.Lesson 57 →
FaaS / ServerlessFunctions as a Service
Run code on events; pay per use; no servers to manage. Example: AWS Lambda resizing images on upload.Lesson 57 →
FinOpsFinancial Operations
Managing and optimising cloud costs. Example: Tagging resources and setting budget alerts.Lesson 57 →
GKE / AKSGoogle Kubernetes Engine / Azure Kubernetes Service
Managed Kubernetes on Google Cloud / Azure. Example: Pokémon GO ran on GKE.Lesson 56 →
IaaSInfrastructure as a Service
Rent virtual machines, disks and networks. Example: AWS EC2, Google Compute Engine.Lesson 57 →
Multi-cloud / Hybrid cloudMulti-cloud / Hybrid cloud
Using several providers / mixing cloud and your own data centre. Example: Some workloads on AWS, some on-premises.Lesson 57 →
PaaSPlatform as a Service
Give your code; the platform runs it. Example: Heroku, Google App Engine.Lesson 57 →
RDSRelational Database Service
AWS managed relational databases. Example: Managed PostgreSQL with backups and failover.Lesson 57 →
RegionRegion
A geographic area with several data centres. Example: ap-south-1 (Mumbai).Lesson 57 →
SaaSSoftware as a Service
Use complete software online. Example: Gmail, Slack, Salesforce.Lesson 57 →
Security groupSecurity group
A virtual firewall for cloud resources. Example: Only allow port 5432 from app servers.Lesson 57 →
Shared responsibility modelShared responsibility model
Provider secures the cloud; you secure what you put in it. Example: You must not make buckets public.Lesson 57 →
Spot instanceSpot instance
Cheap spare cloud capacity that can be taken back. Example: Up to ~90% cheaper for batch jobs.Lesson 57 →
SQS / SNSSimple Queue Service / Simple Notification Service
AWS managed queue / pub-sub. Example: SNS topic fans out to SQS queues.Lesson 35 →
SubnetSubnet
A section of a network's IP range. Example: Public subnet for load balancers, private for apps.Lesson 57 →
VPCVirtual Private Cloud
Your private network inside the cloud. Example: Private subnets for databases.Lesson 57 →

Published by Jainik · Last updated · How these lessons are made

Keep learning

Related series