后端面试练习

REST is an architectural style using resources, HTTP methods, and stateless communication. It is not a query language, CSS framework, or JS library.

HTTP 200 indicates a successful request. 404 means not found, 500 is a server error, and 3xx codes are redirects.

404 means the requested resource could not be found. 200 is success, 401 is unauthorized, and 201 is created.

500 indicates an internal server error on the server side. 200 is success, 400 is a bad request, and 404 is not found.

GET retrieves data and should not change server state. POST creates, PUT updates, and DELETE removes resources.

POST is used to create resources or submit data. GET reads, PUT updates, and DELETE removes resources.

An idempotent operation can be repeated safely with the same effect. GET, PUT, and DELETE are typically idempotent, while POST is not guaranteed.

An endpoint is a URL such as /users that identifies an operation on a resource. It is not a table, log, or class.

JSON is a text format for representing structured data, commonly used in APIs. It is not a database, server, or preprocessor.

Versioning lets APIs evolve without breaking existing clients, often through URL paths or headers. It is not caching, encryption, or log rotation.

Rate limiting controls request frequency per client to protect services. It is not query tuning, response sizing, or memory capping.

A webhook is an HTTP callback sent by a server when an event occurs. It is not a plugin, animation, or database trigger.

An index organizes column values to speed up lookups and filtering. It is not a backup, foreign key, or transaction log, and it can slow writes.

Normalization structures tables to reduce duplication and dependency problems. It is not encryption, sharding, or compression.

A foreign key links rows to a primary key in another table, enforcing relationships. A primary key identifies rows in the same table.

A transaction groups operations with atomicity so they commit or roll back together. It is not a single query, backup, or index.

ACID guarantees atomicity, consistency, isolation, and durability for database transactions. The other sets are not the standard ACID definition.

NoSQL databases use non-relational models such as documents, key-values, or graphs, often for flexible schemas and scale. They still have security and can scale.

MongoDB stores documents in BSON. Redis is a key-value store, while PostgreSQL and MySQL are relational databases.

Redis is an in-memory key-value store often used for caching. MongoDB is document-based, PostgreSQL is relational, and Cassandra is a wide-column store.

Sharding partitions data horizontally across servers to distribute load. It is not encryption, backup, or normalization.

Replication copies data to replica nodes for read scaling and failover. It is not deduplication, indexing, or compression.

A connection pool reuses database connections to reduce setup overhead. It is not a query cache, backup, or firewall.

An ORM maps application objects to database tables and rows. It is not object storage, a queue, or a load balancer.

Migrations manage schema changes with versioned scripts. They are not region copies, log deletion, or restarts.

A query plan shows how the database will execute a query, including index usage and joins. It is not a backup, style guide, or API design.

Caching stores copies of data closer to the caller to reduce latency and load. It is not encryption, backup, or normalization.

Cache invalidation keeps cached data consistent by removing or refreshing entries when the source changes. It is not deletion of servers, encryption, or scaling.

Cache-aside reads from cache first; on a miss, it loads from the database and populates the cache. Write-through writes to cache and database together.

A CDN caches content at edge locations near users to reduce latency. It is not a database, gateway, or broker.

A message queue lets producers and consumers communicate asynchronously. It is not a table, balancer, or cache.

Pub/sub decouples publishers from subscribers through topics or channels. It is not replication, DNS, or caching.

A dead letter queue holds messages that repeatedly fail so they can be inspected later. It is not a general queue or delivery failure state.

Authentication verifies identity, while authorization determines permissions. Encryption and logging are separate concerns.

Authorization decides whether an authenticated user can access a resource. Authentication verifies identity first.

Hashing transforms passwords into irreversible digests, often with salts. It is not encryption, compression, or routing.

A JSON Web Token is a signed token containing claims used for authentication and authorization. It is not a framework, index, or cache key.

TLS encrypts data in transit over HTTPS. It is not a balancer, database, or log format.

Cross-site request forgery tricks a logged-in user into sending unwanted requests. Tokens and same-site cookies help prevent it.

Horizontal scaling adds more machines, while vertical scaling increases resources on one machine. Disk and compression are different.

Vertical scaling increases CPU, memory, or disk on one server. Horizontal scaling adds more servers.

A load balancer spreads incoming requests across healthy servers to improve availability and utilization. It is not encryption, cache, or jobs.

A reverse proxy sits in front of backend servers, handling SSL, routing, and load balancing. It is not a replica, database, or DNS.

Stateless servers do not keep client state between requests, making horizontal scaling easier. Sessions and caches are stateful by nature.

Microservices are small services with clear boundaries that deploy independently. A monolith is a single application.

A monolith packages all features into one application. Microservices split features into separate services.

Availability measures uptime, such as 99.9%. Latency measures speed, and users or data size are separate metrics.

Latency is the time from request to response. Throughput is requests per second, error rate is failures, and memory is resource usage.

Throughput measures capacity such as requests per second. Latency is per-request time, uptime is availability, and hit rate is cache performance.

A circuit breaker opens when a service fails, preventing cascading failures. It is not a lock, switch, or cache.

Graceful degradation serves reduced functionality during failures instead of failing completely. It is not shutdown, restart, or redirect.

An API gateway centralizes routing, auth, rate limiting, and logging for APIs. It is not a pool, broker, or framework.

Eventual consistency means replicas may be briefly different but converge over time. Strong consistency guarantees immediate agreement.

Pagination limits and pages through result sets, often with offset or cursor parameters. It is not encryption, caching, or compression.

Background jobs process work asynchronously, such as emails or image resizing, to keep requests fast. They are not synchronous calls or queries.

Observability uses logs, metrics, and traces to understand and debug systems. It is not a firewall, cache, or balancer.

GET, POST, and PUT are HTTP methods. FETCH is a browser API, not an HTTP method.

Relational, document, key-value, and graph are all database models. They support different data shapes and workloads.

POST is not guaranteed to be idempotent because repeating it can create multiple resources. PUT and DELETE are typically idempotent.

200 means OK, 404 means not found, 500 is an internal server error, and 429 means too many requests.