Skip to main content

Terminology

Common terms and definitions used in our engineering practices.

A​

ACL (Access Control List) Policy that defines which users/services can access which resources.

API (Application Programming Interface) Contract that defines how software components communicate. Usually HTTP-based (REST, GraphQL).

ACID (Atomicity, Consistency, Isolation, Durability) Properties of reliable database transactions. Atomicity = all or nothing. Consistency = valid state. Isolation = concurrent transactions don't interfere. Durability = committed data survives failures.

Authentication Verifying identity: "Are you who you claim to be?" (username + password)

Authorization Granting permissions: "Are you allowed to do this?" (roles + policies)

B​

Blast Radius Scope of systems/users affected by a failure. Small blast radius = failure isolated. Large blast radius = cascading failures.

Blue-Green Deployment Running two identical environments. Deploy to inactive one, switch traffic when ready. Fast rollback by switching back.

C​

CQRS (Command Query Responsibility Segregation) Separating read and write operations into different services/databases. Reads are cheap, writes are reliable.

CI/CD (Continuous Integration / Continuous Deployment) Automated testing and deployment pipeline. Every commit is tested + deployed.

Consensus Multiple parties agreeing on a value. In distributed systems, hard to achieve.

CRUD (Create, Read, Update, Delete) Four basic operations for data persistence.

D​

DRY (Don't Repeat Yourself) Avoid duplicating logic. If you find yourself writing the same code twice, extract it.

Dead Letter Queue Message queue for messages that failed to process. Separates failures from main system.

E​

Eventual Consistency System is consistent eventually, not immediately. Trade-off for availability and partition tolerance.

F​

FIFO (First In, First Out) Queue where first item added is first item removed. Maintains order. (vs LIFO = stack)

Fully Qualified Domain Name (FQDN) Complete domain name: api.example.com (vs api which is relative)

G​

Graceful Degradation Service degrades functionality under load, doesn't crash. E.g., cache miss doesn't break the page.

GRPC High-performance RPC framework. Binary protocol, HTTP/2, strong typing.

H​

Hash One-way function that converts input to fixed-size output. Deterministic (same input = same output).

Idempotent Operation is safe to execute multiple times. Result is the same regardless of execution count. E.g., GET /user/123 is idempotent (doesn't change state).

Immutable Cannot be changed after creation. Benefits: simpler reasoning, thread-safe, cache-friendly.

J​

JWT (JSON Web Token) Self-contained token for authentication. Contains claims (user ID, role, expiry) signed by server. Stateless.

K​

Key-Value Store Simple database: lookup by key, get value. Fast, simple, limited query capability. (Redis, Memcached)

L​

Latency Time to process a request. Measured in milliseconds. Related to throughput but different.

LGTM "Looks Good To Me" - Code review approval.

M​

Mutual TLS (mTLS) TLS where both client and server authenticate each other. Higher security than regular TLS.

N​

NoSQL Non-relational database. Trades ACID for scalability. (MongoDB, DynamoDB, Cassandra)

O​

Observability Can you understand system behavior from outputs (logs, metrics, traces)? Different from monitoring (which tells you if something is wrong).

OAuth2 Standard for delegated authorization. "Let me sign in with GitHub" instead of sharing password.

P​

Partitioning Splitting data/load across multiple servers. By time, geography, or hash of key.

Postmortem Blameless analysis of incident. What happened, why, what's the fix?

P99 Latency 99th percentile latency. 1% of requests are slower than this. Important for user experience.

Q​

Queue FIFO data structure. Add to end, remove from front. Decouples producers from consumers.

R​

RPC (Remote Procedure Call) Calling function on another server as if it were local. (gRPC, XML-RPC, JSON-RPC)

REST (Representational State Transfer) Architectural style for APIs. Resources are identified by URLs, operations are HTTP verbs.

Rollback Revert to previous version. Practice rollbacks before they're needed in emergency.

S​

SLA (Service Level Agreement) Contract with customers about availability. "99.9% uptime" or credits if we miss it.

SLI (Service Level Indicator) Measurable metric of service health. E.g., "error rate < 0.1%".

SLO (Service Level Objective) Target for SLI. E.g., "maintain 99.9% availability over 30 days".

Saturation How full a resource is. CPU at 95% = highly saturated. Measure for scaling.

Sharding Horizontal partitioning. Each shard holds a subset of data. Increases complexity, enables scale.

Synchronous Caller waits for response. Simple but blocks. Prone to cascading failures.

T​

Throughput Number of operations per second. Related to latency but different. High latency ≠ low throughput if parallelized.

TLS (Transport Layer Security) Encryption protocol for HTTPS. Replaces older SSL.

U​

Utilization Percentage of resource being used. 70% CPU = well-utilized. 95% = saturation approaching.

V​

Vertical Scaling Making a single server bigger (more CPU, RAM). Limited by hardware.

W​

Webhook HTTP callback. System A calls HTTP endpoint on System B to notify of event.

Whitelist List of allowed items. "Only these IPs can access this resource".

Z​

Zero-Downtime Deployment Deploy new version without service interruption. (Blue-green, canary, rolling restart)

Documentation​