一、单词
|-------------------------|-------------------------------|----------------------------------|
| Performance | /pərˈfɔːrməns/ | 性能 |
| Optimization | /ˌɑːptɪməˈzeɪʃən/ | 优化 |
| Tuning | /ˈtuːnɪŋ/ | 调优 |
| Bottleneck | /ˈbɑːtəlnek/ | 瓶颈 |
| Response Time | /rɪˈspɑːns taɪm/ | 响应时间 |
| Database Time | /ˈdeɪtəbeɪs taɪm/ | 数据库时间 |
| CPU Time | /siː piː juː taɪm/ | CPU 时间 |
| Wait Time | /weɪt taɪm/ | 等待时间 |
| Runtime Analysis | /ˈrʌntaɪm əˈnæləsɪs/ | 运行时分析 |
| SQL Trace | /es kjuː el treɪs/ | SQL 跟踪 |
| ST05 | /es tiː zɪərəʊ faɪv/ | SQL 跟踪事务代码 |
| SE30 | /es iː θɜːrti/ | 运行时分析事务代码 (旧) |
| SAT | /es eɪ tiː/ | 运行时分析事务代码 (新) |
| STAD | /es tiː eɪ diː/ | 工作负载分析事务代码 |
| ST04 | /es tiː zɪərəʊ fɔːr/ | 数据库性能分析事务代码 |
| DB02 | /diː biː zɪərəʊ tuː/ | 数据库表分析事务代码 |
| Index | /ˈɪndeks/ | 索引 |
| Primary Key | /ˈpraɪmeri kiː/ | 主键 |
| Secondary Index | /ˈsekənderi ˈɪndeks/ | 二级索引 |
| Unique Index | /juˈniːk ˈɪndeks/ | 唯一索引 |
| Composite Index | /kəmˈpɑːzət ˈɪndeks/ | 复合索引 |
| Index Selectivity | /ˈɪndeks səˌlekˈtɪvəti/ | 索引选择性 |
| Full Table Scan | /fʊl ˈteɪbl skæn/ | 全表扫描 |
| Index Scan | /ˈɪndeks skæn/ | 索引扫描 |
| Execution Plan | /ˌeksɪˈkjuːʃən plæn/ | 执行计划 |
| Explain | /ɪkˈspleɪn/ | 解释 (执行计划) |
| Database Statistics | /ˈdeɪtəbeɪs stəˈtɪstɪks/ | 数据库统计信息 |
| Cost-Based Optimization | /kɔːst beɪst ˌɑːptɪməˈzeɪʃən/ | 基于成本的优化 |
| Rule-Based Optimization | /ruːl beɪst ˌɑːptɪməˈzeɪʃən/ | 基于规则的优化 |
| Buffer | /ˈbʌfər/ | 缓冲区 |
| Cache | /kæʃ/ | 缓存 |
| Table Buffer | /ˈteɪbl ˈbʌfər/ | 表缓冲 |
| Buffer Hit Ratio | /ˈbʌfər hɪt ˈreɪʃioʊ/ | 缓冲区命中率 |
| Shared Memory | /ʃerd ˈmeməri/ | 共享内存 |
| Paging | /ˈpeɪdʒɪŋ/ | 分页 |
| Swapping | /ˈswɑːpɪŋ/ | 交换 |
| Open SQL | /ˈoʊpən es kjuː el/ | 开放 SQL |
| Native SQL | /ˈneɪtɪv es kjuː el/ | 原生 SQL |
| Select Statement | /sɪˈlekt ˈsteɪtmənt/ | 查询语句 |
| Update Statement | /ˈʌpdeɪt ˈsteɪtmənt/ | 更新语句 |
| Insert Statement | /ɪnˈsɜːrt ˈsteɪtmənt/ | 插入语句 |
| Delete Statement | /dɪˈliːt ˈsteɪtmənt/ | 删除语句 |
| Modify Statement | /ˈmɑːdɪfaɪ ˈsteɪtmənt/ | 修改语句 |
| For All Entries | /fɔːr ɔːl ˈentrɪz/ | 用于所有条目 (ABAP 语句) |
| Select Single | /sɪˈlekt ˈsɪŋɡl/ | 查询单条记录 |
| Select Distinct | /sɪˈlekt dɪˈstɪŋkt/ | 查询去重记录 |
| Select Ends Elect | /sɪˈlekt endz ɪˈlekt/ | SELECT...ENDSELECT 循环 |
| Nested Loop | /ˈnestɪd luːp/ | 嵌套循环 |
| Nested Select | /ˈnestɪd sɪˈlekt/ | 嵌套查询 |
| Inner Join | /ˈɪnər dʒɔɪn/ | 内连接 |
| Left Outer Join | /left ˈaʊtər dʒɔɪn/ | 左外连接 |
| Right Outer Join | /raɪt ˈaʊtər dʒɔɪn/ | 右外连接 |
| Full Outer Join | /fʊl ˈaʊtər dʒɔɪn/ | 全外连接 |
| Group By | /ɡruːp baɪ/ | 分组 |
| Order By | /ˈɔːrdər baɪ/ | 排序 |
| Having | /ˈhævɪŋ/ | 分组条件 |
| Where Clause | /wer klɔːz/ | WHERE 子句 |
| Aggregate Function | /ˈæɡrɪɡət ˈfʌŋkʃən/ | 聚合函数 |
| Sum | /sʌm/ | 求和 |
| Count | /kaʊnt/ | 计数 |
| Average | /ˈævərɪdʒ/ | 平均值 |
| Max | /mæks/ | 最大值 |
| Min | /mɪn/ | 最小值 |
| Database Lock | /ˈdeɪtəbeɪs lɑːk/ | 数据库锁 |
| Table Lock | /ˈteɪbl lɑːk/ | 表锁 |
| Row Lock | /roʊ lɑːk/ | 行锁 |
| Deadlock | /ˈdedlɑːk/ | 死锁 |
| Enqueue | /ɪnˈkjuː/ | 入队 (加锁) |
| Dequeue | /diːˈkjuː/ | 出队 (解锁) |
| SM12 | /es em twɛlv/ | 锁监控事务代码 |
| Batch Input | /bætʃ ˈɪnpʊt/ | 批输入 |
| BDC | /biː diː siː/ | 批数据通信 (Batch Data Communication) |
| LSMW | /el es em ˈdʌbəljuː/ | 数据迁移工作台 |
| BAPI | /ˈbɑːpi/ | 业务应用程序编程接口 |
| Data Volume | /ˈdeɪtə ˈvɑːljuːm/ | 数据量 |
| Large Data | /lɑːrdʒ ˈdeɪtə/ | 大数据 |
| Mass Processing | /mæs ˈprɑːsesɪŋ/ | 批量处理 |
| Parallel Processing | /ˈpærəlel ˈprɑːsesɪŋ/ | 并行处理 |
| Background Job | /ˈbækɡraʊnd dʒoʊb/ | 后台作业 |
| SM37 | /es em θɜːrti ˈsevən/ | 后台作业监控事务代码 |
| Update Task | /ˈʌpdeɪt tæsk/ | 更新任务 |
| Asynchronous Update | /eɪˈsɪŋkrənəs ʌpˈdeɪt/ | 异步更新 |
| Synchronous Update | /ˈsɪŋkrənəs ʌpˈdeɪt/ | 同步更新 |
| Variant | /ˈveriənt/ | 变式 |
| Authorization Check | /ˌɔːθərəˈzeɪʃən tʃek/ | 权限检查 |
| Table Partitioning | /ˈteɪbl pɑːrˈtɪʃənɪŋ/ | 表分区 |
| Table Compression | /ˈteɪbl kəmˈpreʃən/ | 表压缩 |
| Database Hint | /ˈdeɪtəbeɪs hɪnt/ | 数据库提示 |
| Client | /ˈklaɪənt/ | 客户端 |
| Mandt | /mɑːnt/ | 客户端字段 (MANDT) |
| Logical Database | /ˈlɑːdʒɪkl ˈdeɪtəbeɪs/ | 逻辑数据库 |
| View | /vjuː/ | 视图 |
| Database View | /ˈdeɪtəbeɪs vjuː/ | 数据库视图 |
| Projection View | /prəˈdʒekʃən vjuː/ | 投影视图 |
| Maintenance View | /ˈmeɪntənəns vjuː/ | 维护视图 |
| Help View | /help vjuː/ | 帮助视图 |
| Pooled Table | /puːld ˈteɪbl/ | 池表 |
| Cluster Table | /ˈklʌstər ˈteɪbl/ | 簇表 |
| Transparent Table | /trænsˈpærənt ˈteɪbl/ | 透明表 |
| Archive | /ˈɑːrkaɪv/ | 归档 |
| Data Aging | /ˈdeɪtə ˈeɪdʒɪŋ/ | 数据老化 (HANA 特性) |
| AMDP | /eɪ em diː piː/ | ABAP 托管数据库过程 |
| CDS View | /siː diː es vjuː/ | 核心数据服务视图 |
| HANA | /ˈhɑːnə/ | SAP HANA 数据库 |
| In-Memory Database | /ɪn ˈmeməri ˈdeɪtəbeɪs/ | 内存数据库 |
二、短语
|-------------------------------------------------------|------------------------------|
| Optimize the SELECT statement using indexes | "使用索引优化 SELECT 语句" |
| Avoid nested SELECT statements inside loops | "避免在循环中使用嵌套 SELECT 语句" |
| Use FOR ALL ENTRIES to reduce database access | "使用 FOR ALL ENTRIES 减少数据库访问" |
| Check the execution plan using transaction ST05 | "使用事务代码 ST05 检查执行计划" |
| Analyze the runtime with transaction SE30 | "使用事务代码 SE30 分析运行时" |
| Create secondary indexes on frequently queried fields | "在频繁查询的字段上创建二级索引" |
| Implement parallel processing for large data volumes | "对大数据量实施并行处理" |
| Use aggregate functions to minimize data transfer | "使用聚合函数最小化数据传输" |
| Check the database buffer hit ratio in ST04 | "在 ST04 中检查数据库缓冲区命中率" |
| Handle database locks properly to avoid deadlocks | "妥善处理数据库锁以避免死锁" |
三、对话
Q: What is the most common performance issue you have encountered?
A: The most common issue is SELECT statements inside loops. For example, reading header data and then for each header, selecting item details individually. This causes thousands of database round trips. I always refactor this by using FOR ALL ENTRIES or a single JOIN to fetch all required data in one go. In one project, this optimization reduced a report's runtime from 2 hours to 15 minutes.
问：你遇到的最常见的性能问题是什么？
答：最常见的问题是在循环中使用 SELECT 语句。例如，读取抬头数据，然后为每个抬头单独选择行项目明细。这会导致数千次数据库往返。我总是通过使用 FOR ALL ENTRIES 或单个 JOIN 一次性获取所有所需数据来重构这个问题。在一个项目中，这次优化将一个报表的运行时间从 2 小时减少到了 15 分钟。
Q: Can you explain the difference between using JOIN and FOR ALL ENTRIES?A: JOIN combines data from multiple tables in a single database operation, which is very efficient. FOR ALL ENTRIES selects data based on an internal table, generating multiple SELECT statements. JOIN is preferred when the relationship is clear and all data comes from the same database. FOR ALL ENTRIES is useful when you need to avoid duplicates or when the selection criteria is complex. However, always check that the internal table is not empty when using FOR ALL ENTRIES, as it would bypass all WHERE conditions.
问：你能解释一下使用 JOIN 和 FOR ALL ENTRIES 的区别吗？
答：JOIN 在单个数据库操作中组合来自多个表的数据，这是非常高效的。FOR ALL ENTRIES 基于内表选择数据，生成多个 SELECT 语句。当关系清晰且所有数据来自同一个数据库时，首选 JOIN。当需要避免重复或选择条件复杂时，FOR ALL ENTRIES 很有用。但是，使用 FOR ALL ENTRIES 时，始终要检查内表不为空，否则会绕过所有 WHERE 条件。
Q: How do you create an index and when do you know you need one?A: I create indexes using transaction SE11. I know I need an index when I see full table scans in the ST05 execution plan for frequently executed queries. I look at the WHERE conditions in slow-running SELECT statements. The index should include the fields used in WHERE clauses, especially equality conditions. I also consider the selectivity of the index - fields with many distinct values make better indexes. However, I avoid creating too many indexes as they slow down update operations.
问：你如何创建索引，以及如何知道何时需要创建索引？
答：我使用事务代码 SE11 创建索引。当我在 ST05 执行计划中看到频繁执行的查询进行全表扫描时，我就知道需要索引。我查看运行缓慢的 SELECT 语句中的 WHERE 条件。索引应包含 WHERE 子句中使用的字段，特别是等值条件。我还会考虑索引的选择性 - 具有许多不同值的字段能构成更好的索引。但是，我避免创建太多索引，因为它们会拖慢更新操作。
Q: What is the purpose of database statistics and how often should they be updated?A: Database statistics help the query optimizer choose the most efficient execution plan. They contain information about table sizes, data distribution, and index selectivity. Without updated statistics, the optimizer might make poor decisions. Statistics should be updated regularly, especially after significant data changes. In SAP systems, we typically run DBACOCKPIT or have background jobs scheduled to update statistics weekly. For critical tables with frequent data changes, more frequent updates might be necessary.
问：数据库统计信息的目的是什么，应该多久更新一次？
答：数据库统计信息帮助查询优化器选择最高效的执行计划。它们包含关于表大小、数据分布和索引选择性的信息。如果没有更新的统计信息，优化器可能会做出错误的决策。统计信息应定期更新，特别是在大量数据变更之后。在 SAP 系统中，我们通常运行 DBACOCKPIT 或安排后台作业每周更新统计信息。对于数据变更频繁的关键表，可能需要更频繁的更新。
Q: How do you handle large data volumes in reporting without causing performance issues?A: For large data volumes, I implement several strategies. First, I use parallel processing by splitting the data into packages and processing them simultaneously using asynchronous RFC. Second, I process data in chunks to avoid memory issues, using the UP TO...OFFSET technique. Third, I consider using database features like AMDP for HANA to push logic down to the database layer. Fourth, I ensure that reports have proper selection criteria to limit data. Finally, for very large reports, I suggest background processing with results stored in a table for quick access.
问：你如何处理报表中的大数据量而不导致性能问题？
答：对于大数据量，我实施多种策略。首先，我使用并行处理，将数据拆分成多个包并使用异步 RFC 同时处理。其次，我使用 UP TO...OFFSET 技术分块处理数据以避免内存问题。第三，我考虑使用 HANA 的 AMDP 等数据库特性，将逻辑下推到数据库层。第四，我确保报表有适当的选择条件来限制数据。最后，对于非常大的报表，我建议进行后台处理，并将结果存储在表中以便快速访问。
Q: Have you ever dealt with a deadlock situation? How did you resolve it?A: Yes, I've encountered deadlocks in high-volume batch processing. In one case, two programs were updating the same tables but in different orders. I analyzed the issue using SM12 and ST05 traces. The solution was to standardize the lock sequence - ensuring all programs lock objects in the same order. I also implemented retry logic with exponential backoff for programs that couldn't be changed. Additionally, I reduced transaction scope to lock only necessary records and for the shortest time possible, which minimized contention.
问：你处理过死锁情况吗？你是如何解决的？
答：是的，我在高容量批处理中遇到过死锁。在一个案例中，两个程序更新相同的表，但顺序不同。我使用 SM12 和 ST05 跟踪分析了问题。解决方案是标准化锁顺序 - 确保所有程序以相同的顺序锁定对象。我还为无法更改的程序实现了指数退避的重试逻辑。此外，我缩小了事务范围，只锁定必要的记录，并尽可能缩短锁定时间，这最大限度地减少了争用
Q: What is the impact of SELECT * on performance and why?A: SELECT * retrieves all fields from a table, which is inefficient. It transfers unnecessary data from the database to the application server, increasing network load and memory consumption. It can also prevent the use of covering indexes, where all needed fields are in the index itself. Always specify only the fields you actually need. For example, instead of SELECT * FROM MARA, use SELECT MATNR, MTART FROM MARA. This small change can significantly improve performance, especially for large tables.
问：SELECT * 对性能有什么影响，为什么？
答：SELECT * 检索表中的所有字段，这是低效的。它将不必要的数据从数据库传输到应用服务器，增加了网络负载和内存消耗。它还可能阻止使用覆盖索引，即所有需要的字段都在索引本身中。始终只指定你实际需要的字段。例如，不要使用 SELECT * FROM MARA，而应使用 SELECT MATNR
Q: How do you monitor system performance in a live environment?A: I regularly use several monitoring tools. ST03N shows workload analysis and response time trends. ST04 monitors database performance including buffer hit ratios. AL08 shows current users and processes. For long-running jobs, I check SM37 logs. I also set up alerts in Solution Manager for critical thresholds. When users report issues, I first check STAD for the specific transaction's runtime profile. Regular monitoring helps identify potential issues before they impact users.
问：你在生产环境中如何监控系统性能？
答：我经常使用多种监控工具。ST03N 显示工作负载分析和响应时间趋势。ST04 监控数据库性能，包括缓冲区命中率。AL08 显示当前用户和进程。对于长时间运行的作业，我会检查 SM37 日志。我还在 Solution Manager 中设置关键阈值的警报。当用户报告问题时，我首先检查特定事务的 STAD 运行时概况。定期监控有助于在潜在问题影响用户之前发现它们。
Q: What are your top three tips for writing high-performance ABAP code?A: First, minimize database accesses. Instead of SELECT inside loops, use JOINs or FOR ALL ENTRIES to get all data at once. Second, use appropriate internal table types - sorted tables for large lookups with binary search instead of linear search in standard tables. Third, be selective about data: use WHERE clauses effectively, select only needed fields, and avoid SELECT DISTINCT by removing duplicates in ABAP using DELETE ADJACENT DUPLICATES after sorting. These three principles have consistently helped me deliver performant applications.
问：你编写高性能 ABAP 代码的三个最重要的技巧是什么？
答：首先，最小化数据库访问。不要在循环中使用 SELECT，而应使用 JOIN 或 FOR ALL ENTRIES 一次性获取所有数据。第二，使用合适的内表类型 - 对于大型查找，使用排序表和二分查找，而不是在标准表中进行线性查找。第三，对数据要有选择性：有效使用 WHERE 子句，只选择需要的字段，通过在排序后使用 DELETE ADJACENT DUPLICATES 在 ABAP 中去除重复项来避免 SELECT DISTINCT。这三个原则一直帮助我交付高性能的应用程序