先背你每天用的领域，不要按字母表背。 每个词都是你工作中会遇到的高频词，不是考试词表。
面试追问地图
| 你要 | 建议先背 | 大约词汇量 |
|---|---|---|
| 看懂技术文档 | Java/数据库 + 通用动词 | ~200 词 |
| 写英文邮件/PR | 通用动词 + 工程实践 | +100 词 |
| 技术讨论 | 分布式/系统设计 + 网络 | +150 词 |
| 技术面试 | 全领域 + 算法术语 | +150 词 |
一、通用高频动词
这些词是你写英文邮件、PR 描述、代码注释时最常用的。
| 英文 | 中文 | 常见搭配 | 例句 |
|---|---|---|---|
| implement | 实现 | implement a feature / interface | We need to implement the caching layer. |
| resolve | 解决 | resolve an issue / conflict | This PR resolves #1234. |
| optimize | 优化 | optimize performance / query | The query was optimized by adding an index. |
| refactor | 重构 | refactor the code / module | Refactored the payment service to reduce coupling. |
| deploy | 部署 | deploy to production / staging | The fix was deployed to production at 3pm. |
| trigger | 触发 | trigger an event / build | Pushing to main triggers the CI pipeline. |
| invoke | 调用 | invoke a method / API | The service invokes the payment gateway. |
| propagate | 传播 | propagate the error / context | The trace context is propagated across services. |
| aggregate | 聚合 | aggregate data / results | The metrics are aggregated every minute. |
| persist | 持久化 | persist data / state | The session data is persisted to Redis. |
| retrieve | 检索 | retrieve data / records | The API retrieves user profiles from the cache. |
| validate | 校验 | validate input / token | The interceptor validates the JWT token. |
| serialize | 序列化 | serialize object / response | The response is serialized to JSON. |
| deserialize | 反序列化 | deserialize payload / request body | The controller deserializes the request body. |
| replicate | 复制 | replicate data / database | The data is replicated across three nodes. |
| evict | 驱逐 | evict cache / entry | The cache evicts the least recently used entries. |
| throttle | 限流 | throttle requests / rate | The API gateway throttles requests exceeding the limit. |
| degrade | 降级 | degrade gracefully / service | The service degrades to a fallback response. |
| bootstrap | 初始化 | bootstrap the application / context | Spring bootstraps the application context on startup. |
| orchestrate | 编排 | orchestrate services / workflow | The saga orchestrates the distributed transaction. |
二、通用高频名词
| 英文 | 中文 | 常见搭配 | 例句 |
|---|---|---|---|
| latency | 延迟 | reduce latency / network latency | The cache significantly reduces latency. |
| throughput | 吞吐量 | increase throughput / high throughput | Kafka is designed for high throughput. |
| overhead | 开销 | memory overhead / performance overhead | Each thread adds a small overhead. |
| bottleneck | 瓶颈 | performance bottleneck / identify the bottleneck | The database connection pool is the bottleneck. |
| trade-off | 权衡 | a trade-off between A and B | There’s a trade-off between consistency and availability. |
| constraint | 约束 | a design constraint / under constraint | The system must operate under memory constraints. |
| redundancy | 冗余 | data redundancy / add redundancy | Redundancy is achieved through replication. |
| resilience | 弹性 | system resilience / build resilience | Circuit breakers improve system resilience. |
| consistency | 一致性 | strong consistency / eventual consistency | The system guarantees eventual consistency. |
| availability | 可用性 | high availability / ensure availability | The cluster provides 99.99% availability. |
| scalability | 可扩展性 | horizontal scalability / scale out | The architecture supports horizontal scalability. |
| reliability | 可靠性 | system reliability / improve reliability | Retries and timeouts improve reliability. |
| concurrency | 并发 | handle concurrency / concurrency control | The lock manager handles concurrency. |
| dependency | 依赖 | a dependency on / manage dependencies | The service has a dependency on the user service. |
| contract | 契约 | API contract / break the contract | The API contract is defined in the OpenAPI spec. |
| payload | 载荷 | request payload / message payload | The payload is compressed before sending. |
| schema | 模式 | database schema / schema migration | The schema migration is applied by Flyway. |
| workload | 工作负载 | read-heavy workload / mixed workload | The database handles a read-heavy workload. |
三、Java / JVM
| 英文 | 中文 | 说明 |
|---|---|---|
| heap / stack | 堆 / 栈 | JVM memory areas |
| garbage collection (GC) | 垃圾回收 | automatic memory management |
| classloader | 类加载器 | loads class files into JVM |
| bytecode | 字节码 | JVM instruction set |
| just-in-time (JIT) compilation | 即时编译 | compiles bytecode to native code |
| thread pool | 线程池 | manages a pool of worker threads |
| lock contention | 锁竞争 | multiple threads competing for a lock |
| deadlock | 死锁 | two threads waiting for each other’s locks |
| race condition | 竞态条件 | incorrect behavior due to timing |
| thread-safe | 线程安全 | safe for concurrent use |
| immutable | 不可变 | state cannot change after creation |
| serialization | 序列化 | converting object to byte stream |
| reflection | 反射 | inspecting/modifying runtime behavior |
| annotation | 注解 | metadata for classes/methods |
| aspect-oriented programming (AOP) | 面向切面编程 | cross-cutting concerns |
| dependency injection (DI) | 依赖注入 | injecting dependencies into objects |
| inversion of control (IoC) | 控制反转 | framework calls your code |
| proxy | 代理 | an object representing another object |
| bean | Bean | Spring-managed object |
| autowiring | 自动装配 | automatic dependency resolution |
| connection pool | 连接池 | pool of reusable connections |
| thread dump | 线程转储 | snapshot of all thread states |
| heap dump | 堆转储 | snapshot of all objects in heap |
| garbage collector | 垃圾收集器 | e.g. G1, ZGC, Shenandoah |
| stop-the-world (STW) | 全局暂停 | all application threads are paused |
| memory leak | 内存泄漏 | objects not garbage collected |
| off-heap | 堆外 | memory outside the JVM heap |
| volatile | 可见性 | guarantees visibility of changes |
| happens-before | 先行发生 | ordering guarantee in JMM |
四、数据库
MySQL
| 英文 | 中文 | 说明 |
|---|---|---|
| index | 索引 | data structure for fast lookup |
| primary key | 主键 | unique identifier for a row |
| foreign key | 外键 | reference to another table |
| composite index | 联合索引 | index on multiple columns |
| covering index | 覆盖索引 | index that satisfies a query without lookup |
| clustered index | 聚簇索引 | data rows are stored in index order |
| B+ tree | B+ 树 | balanced tree for disk-based storage |
| transaction | 事务 | a unit of work |
| ACID | 原子性/一致性/隔离性/持久性 | transaction properties |
| commit | 提交 | make changes permanent |
| rollback | 回滚 | undo changes |
| isolation level | 隔离级别 | how transactions see each other |
| dirty read | 脏读 | reading uncommitted data |
| phantom read | 幻读 | same query returns different rows |
| MVCC | 多版本并发控制 | multi-version concurrency control |
| undo log | 回滚日志 | for rolling back transactions |
| redo log | 重做日志 | for crash recovery |
| binlog | 二进制日志 | for replication and point-in-time recovery |
| slow query | 慢查询 | query exceeding a time threshold |
| execution plan | 执行计划 | how the database executes a query |
| full table scan | 全表扫描 | reading every row |
| table lock | 表锁 | lock on the entire table |
| row lock | 行锁 | lock on a specific row |
| gap lock | 间隙锁 | lock on the gap between records |
| deadlock | 死锁 | two transactions waiting for each other |
| master-slave | 主从 | replication topology |
| sharding | 分库分表 | splitting data across multiple databases |
| partition | 分区 | dividing a table into smaller pieces |
Redis
| 英文 | 中文 | 说明 |
|---|---|---|
| in-memory | 内存型 | data stored in RAM |
| key-value store | 键值存储 | stores data as key-value pairs |
| data structure | 数据结构 | string, list, set, hash, sorted set, stream, etc. |
| expiration | 过期 | automatic key deletion |
| TTL | 生存时间 | time-to-live for a key |
| eviction | 驱逐 | removing keys when memory is full |
| persistence | 持久化 | saving data to disk |
| RDB | 快照 | point-in-time snapshot |
| AOF | 追加文件 | append-only file for logging writes |
| pipeline | 管道 | batching commands |
| pub/sub | 发布/订阅 | messaging pattern |
| Lua script | Lua 脚本 | atomic script execution |
| sentinel | 哨兵 | high availability for Redis |
| cluster | 集群 | distributed Redis |
| hash slot | 哈希槽 | key distribution in cluster mode |
| cache penetration | 缓存穿透 | querying for non-existent data |
| cache breakdown | 缓存击穿 | a hot key expires |
| cache avalanche | 缓存雪崩 | many keys expire at once |
| bloom filter | 布隆过滤器 | probabilistic set membership |
| bitmap | 位图 | bit-level operations |
| hyperloglog | 超对数 | cardinality estimation |
| geo | 地理位置 | geospatial data |
| stream | 流 | append-only log for message streaming |
五、分布式系统
| 英文 | 中文 | 说明 |
|---|---|---|
| distributed system | 分布式系统 | system running on multiple nodes |
| consensus | 共识 | agreeing on a single value |
| leader election | 选主 | choosing a coordinator |
| replication | 复制 | copying data across nodes |
| sharding | 分片 | partitioning data across nodes |
| failover | 故障转移 | switching to a backup |
| split-brain | 脑裂 | cluster splits into two parts |
| quorum | 法定人数 | majority needed for decision |
| heartbeat | 心跳 | periodic signal for liveness |
| gossip protocol | 流言协议 | epidemic information dissemination |
| eventual consistency | 最终一致性 | data eventually consistent |
| strong consistency | 强一致性 | all nodes see the same data |
| CAP theorem | CAP 定理 | consistency / availability / partition tolerance |
| idempotency | 幂等 | same operation, same result |
| exactly-once | 精确一次 | delivery guarantee |
| at-least-once | 至少一次 | delivery guarantee |
| at-most-once | 最多一次 | delivery guarantee |
| circuit breaker | 熔断器 | stop calling a failing service |
| bulkhead | 舱壁 | isolate failures |
| rate limiting | 限流 | control request rate |
| backpressure | 背压 | consumer signals producer to slow down |
| retry | 重试 | try again on failure |
| timeout | 超时 | give up after a duration |
| graceful degradation | 优雅降级 | reduce functionality gracefully |
| fault tolerance | 容错 | system continues despite failures |
| eventual consistency | 最终一致性 | data becomes consistent over time |
六、系统设计
| 英文 | 中文 | 说明 |
|---|---|---|
| load balancer | 负载均衡器 | distributes traffic |
| reverse proxy | 反向代理 | sits in front of servers |
| API gateway | API 网关 | single entry point |
| CDN | 内容分发网络 | content delivery network |
| message queue | 消息队列 | async communication |
| publish-subscribe | 发布订阅 | messaging pattern |
| producer / consumer | 生产者 / 消费者 | messaging roles |
| stateless / stateful | 无状态 / 有状态 | server design |
| horizontal scaling | 水平扩展 | adding more machines |
| vertical scaling | 垂直扩展 | upgrading a single machine |
| caching layer | 缓存层 | e.g., Redis, Memcached |
| database sharding | 数据库分片 | splitting data |
| read replica | 只读副本 | read-only copy |
| write-ahead log (WAL) | 预写日志 | log before modifying |
| leader / follower | 主节点 / 从节点 | replication roles |
| consistent hashing | 一致性哈希 | distributing keys with minimal reshuffling |
| bloom filter | 布隆过滤器 | probabilistic set membership |
| token bucket | 令牌桶 | rate limiting algorithm |
| leaky bucket | 漏桶 | rate limiting algorithm |
| sliding window | 滑动窗口 | time-based rate limiting |
| short URL | 短链接 | URL shortening |
| idempotency key | 幂等键 | prevent duplicate operations |
| event sourcing | 事件溯源 | store events, not state |
| CQRS | 命令查询职责分离 | command query responsibility segregation |
| saga | 长事务 | distributed transaction pattern |
| two-phase commit (2PC) | 两阶段提交 | distributed commit protocol |
七、网络
| 英文 | 中文 | 说明 |
|---|---|---|
| TCP / UDP | 传输控制协议 / 用户数据报协议 | transport layer protocols |
| handshake | 握手 | connection establishment |
| three-way handshake | 三次握手 | TCP connection setup |
| four-way挥手 | 四次挥手 | TCP connection termination |
| congestion control | 拥塞控制 | managing network congestion |
| flow control | 流量控制 | managing sender rate |
| sliding window | 滑动窗口 | flow control mechanism |
| acknowledgment (ACK) | 确认 | packet received |
| retransmission | 重传 | resending lost packets |
| timeout | 超时 | waiting time exceeded |
| keep-alive | 保活 | maintaining idle connection |
| multiplexing | 多路复用 | multiple streams over one connection |
| DNS | 域名系统 | domain name resolution |
| round-robin | 轮询 | distributing requests in turn |
| proxy | 代理 | intermediary server |
| firewall | 防火墙 | security barrier |
| subnet | 子网 | network subdivision |
| latency | 延迟 | time for data to travel |
| bandwidth | 带宽 | data transfer capacity |
| packet loss | 丢包 | packets not arriving |
| HTTP / HTTPS | 超文本传输协议 | application layer protocols |
| REST | 表述性状态转移 | representational state transfer |
| WebSocket | 全双工通信 | persistent bidirectional connection |
| gRPC | gRPC | high-performance RPC framework |
| TLS | 传输层安全 | transport layer security |
| certificate | 证书 | identity verification |
| load balancing | 负载均衡 | distributing traffic |
八、操作系统
| 英文 | 中文 | 说明 |
|---|---|---|
| process | 进程 | running program |
| thread | 线程 | lightweight unit of execution |
| context switch | 上下文切换 | switching between processes/threads |
| scheduler | 调度器 | decides which thread runs |
| preemption | 抢占 | forcibly taking CPU |
| time slice | 时间片 | CPU time quantum |
| virtual memory | 虚拟内存 | abstracting physical memory |
| page fault | 缺页 | accessing a page not in memory |
| page table | 页表 | maps virtual to physical addresses |
| TLB | 快表 | translation lookaside buffer |
| swap | 交换 | moving pages to disk |
| kernel | 内核 | core of the OS |
| system call | 系统调用 | requesting kernel service |
| user space / kernel space | 用户态 / 内核态 | privilege levels |
| interrupt | 中断 | signal to the processor |
| deadlock | 死锁 | circular waiting for resources |
| mutex | 互斥锁 | mutual exclusion |
| semaphore | 信号量 | counting synchronization |
| spinlock | 自旋锁 | busy-waiting lock |
| memory-mapped file | 内存映射文件 | file mapped to memory |
| zero-copy | 零拷贝 | avoid copying data |
| buffer | 缓冲区 | temporary storage |
| DMA | 直接内存访问 | direct memory access |
| I/O multiplexing | I/O 多路复用 | monitoring multiple I/O streams |
| epoll | epoll | Linux I/O event notification |
| page cache | 页缓存 | cached file pages |
| NUMA | 非统一内存访问 | non-uniform memory access |
九、算法与数据结构
| 英文 | 中文 | 说明 |
|---|---|---|
| array | 数组 | contiguous memory |
| linked list | 链表 | nodes linked by pointers |
| stack | 栈 | LIFO |
| queue | 队列 | FIFO |
| hash table | 哈希表 | key-value mapping |
| tree | 树 | hierarchical structure |
| binary tree | 二叉树 | each node has at most two children |
| binary search tree (BST) | 二叉搜索树 | left < root < right |
| balanced tree | 平衡树 | height-balanced |
| heap | 堆 | complete binary tree with heap property |
| graph | 图 | nodes and edges |
| trie | 前缀树 | prefix-based tree |
| recursion | 递归 | function calling itself |
| iteration | 迭代 | looping |
| backtracking | 回溯 | exploring all possibilities |
| dynamic programming (DP) | 动态规划 | overlapping subproblems |
| greedy algorithm | 贪心算法 | locally optimal choice |
| sliding window | 滑动窗口 | subarray technique |
| two pointers | 双指针 | converging / diverging pointers |
| binary search | 二分查找 | divide and conquer on sorted data |
| breadth-first search (BFS) | 广度优先搜索 | level by level |
| depth-first search (DFS) | 深度优先搜索 | go deep first |
| topological sort | 拓扑排序 | ordering in a DAG |
| shortest path | 最短路径 | Dijkstra / Bellman-Ford / Floyd-Warshall |
| minimum spanning tree | 最小生成树 | Kruskal / Prim |
| time complexity | 时间复杂度 | Big O notation |
| space complexity | 空间复杂度 | memory usage |
| best / worst / average case | 最好/最坏/平均情况 | performance analysis |
| divide and conquer | 分治 | split, solve, merge |
| memoization | 记忆化 | caching results of function calls |
| adjacency list | 邻接表 | graph representation |
| adjacency matrix | 邻接矩阵 | graph representation |
| DAG | 有向无环图 | directed acyclic graph |
| topological order | 拓扑序 | linear ordering of vertices |
十、工程实践
| 英文 | 中文 | 说明 |
|---|---|---|
| version control | 版本控制 | Git, SVN |
| repository | 仓库 | code storage |
| commit | 提交 | save changes |
| branch | 分支 | independent line of development |
| merge | 合并 | combining branches |
| rebase | 变基 | replaying commits |
| pull request (PR) | 拉取请求 | code review request |
| code review | 代码审查 | reviewing code |
| continuous integration (CI) | 持续集成 | automated build and test |
| continuous deployment (CD) | 持续部署 | automated deployment |
| pipeline | 流水线 | CI/CD workflow |
| artifact | 制品 | build output |
| rollback | 回滚 | reverting deployment |
| blue-green deployment | 蓝绿部署 | two environments |
| canary release | 金丝雀发布 | gradual rollout |
| feature flag | 特性开关 | toggle feature on/off |
| unit test | 单元测试 | testing individual units |
| integration test | 集成测试 | testing component interaction |
| end-to-end test (E2E) | 端到端测试 | testing full workflow |
| regression | 回归 | previously working feature breaks |
| mock / stub | 模拟 / 桩 | test doubles |
| assertion | 断言 | verify expected outcome |
| code coverage | 代码覆盖率 | percentage of code tested |
| linting | 代码检查 | static analysis |
| logging | 日志 | recording events |
| monitoring | 监控 | observing system health |
| alerting | 告警 | notifying about issues |
| metrics | 指标 | quantitative measurements |
| dashboard | 仪表盘 | visual display of metrics |
| on-call | 值班 | being available for incidents |
| postmortem | 事后复盘 | incident analysis |
| runbook | 操作手册 | operational procedures |
| service level agreement (SLA) | 服务等级协议 | commitment to customers |
| service level objective (SLO) | 服务等级目标 | internal target |
| error budget | 错误预算 | allowed failure rate |
| toil | 重复劳动 | manual repetitive work |
| reliability | 可靠性 | system works correctly |
| observability | 可观测性 | understanding system internals |
| tracing | 链路追踪 | tracking request flow |
| profiling | 性能分析 | analyzing performance |
常见追问
这么多词，怎么背？
不要背，要”遇到-查-用”。 每天读官方文档，遇到不认识的词用沙拉查词划一下，一键导入 Anki。工作中写邮件/PR 时，对照这个表查词。一个月后，你每天用的词自然就记住了。
音标要背吗？
不需要。 技术英语的发音重点是”让对方听懂”，不是”发音完美”。遇到不确定的发音，用 Google 翻译的发音按钮听一遍，模仿即可。重点纠正几个影响理解的音（如 th / v-w / 长短元音），见 技术听说。
技术面试中英语不好怎么办？
代码 > 英语。 如果说不清楚，就在白板上写出来。准备一个”技术面试英语表达”清单，提前背熟（见 面试英语）。面试官看重的是你的技术能力，不是英语水平。