scheduler ■ case study: batching client requests ● Memory control ■ case study: transparent huge pages ● GC Related ■ case study: GC sweep caused latency jitter ■ case study: Lock and NUMA aware Agenda ■ each process has its virtual address space ○ Physical memory ■ OS divides physical memory into pages ○ memory mapping ■ page fault, built the virtual-physical mapping ○ RSS (resident set size) ■ Held Result in better memory access performance Background: THP (transparent huge pages) Case study: THP (transparent huge pages) Description ● The TiDB server memory footprint is abnormal ● The memory

MySQL. Note: This page describes general differences between MySQL and TiDB. See the dedicated pages for Security and Pessimistic Transaction Mode compatibility. 2.6.1 Unsupported features • Stored listed per page. In addition, the return value contains pagination information such as offset, total pages, and whether the results are sorted. 176 { "content": [ { "coins": 200, "goods": 10, "id": 1 unique index to sort results and the offset keyword in the LIMIT clause to split pages by a specified row count. Then the pages are wrapped into independent transactions to achieve flexible paging updates

MySQL. Note: This page refers to general differences between MySQL and TiDB. Refer to the dedicated pages for Security and Pessimistic Transaction Model compatibility. 2.6.1 Unsupported features • Stored (Transparent Huge Pages). The memory access pattern of databases tends to be sparse rather than consecutive. If the high-level memory fragmentation is serious, higher latency will occur when THP pages are allocated programs, you can use the numactl command for the CPU binding. For detailed usage, see the Linux manual pages. For network interface card (NIC) interrupts, see tune network. 9.1.1.2.4 Memory—transparent huge

MySQL. Note: This page refers to general differences between MySQL and TiDB. Refer to the dedicated pages for Security and Pessimistic Transaction Mode compatibility. 2.6.1 Unsupported features • Stored (Transparent Huge Pages). The memory access pattern of databases tends to be sparse rather than consecutive. If the high-level memory fragmentation is serious, higher latency will occur when THP pages are allocated programs, you can use the numactl command for the CPU binding. For detailed usage, see the Linux manual pages. For network interface card (NIC) interrupts, see tune network. 9.1.1.2.4 Memory—transparent huge

MySQL. Note: This page refers to general differences between MySQL and TiDB. Refer to the dedicated pages for Security and Pessimistic Transaction Model compatibility. 2.6.1 Unsupported features • Stored (Transparent Huge Pages). The memory access pattern of databases tends to be sparse rather than consecutive. If the high-level memory fragmentation is serious, higher latency will occur when THP pages are allocated programs, you can use the numactl command for the CPU binding. For detailed usage, see the Linux manual pages. For network interface card (NIC) interrupts, see tune network. 9.1.1.2.4 Memory—transparent huge

MySQL. Note: This page describes general differences between MySQL and TiDB. See the dedicated pages for Security and Pessimistic Transaction Mode compatibility. 95 You can try out TiDB features on unique index to sort results and the offset keyword in the LIMIT clause to split pages by a specified row count. Then the pages are wrapped into independent transactions to achieve flexible paging updates of a book is not required on the book lists page but only required on the book details and order pages. Stale Read can be used to improve throughout of the application. 4.7.8.2.2 Statement level To query

MySQL. Note: This page describes general differences between MySQL and TiDB. See the dedicated pages for Security and Pessimistic Transaction Mode compatibility. 2.6.1 Unsupported features • Stored (Transparent Huge Pages). The memory access pattern of databases tends to be sparse rather than consecutive. If the high-level memory fragmentation is serious, higher latency will occur when THP pages are allocated programs, you can use the numactl command for the CPU binding. For detailed usage, see the Linux manual pages. For network interface card (NIC) interrupts, see tune network. 9.1.1.2.4 Memory—transparent huge

compatibility with MySQL in the areas of security and pes- simistic transaction mode, refer to the dedicated pages on Security and Pes- simistic Transaction Mode. You can try out TiDB features on TiDB Playground unique index to sort results and the offset keyword in the LIMIT clause to split pages by a specified row count. Then the pages are wrapped into independent transactions to achieve flexible paging updates of a book is not required on the book lists page but only required on the book details and order pages. Stale Read can be used to improve throughout of the application. 4.7.8.2.2 Statement level To query

compatibility with MySQL in the areas of security and pes- simistic transaction mode, refer to the dedicated pages on Security and Pes- simistic Transaction Mode. You can try out TiDB features on TiDB Playground unique index to sort results and the offset keyword in the LIMIT clause to split pages by a specified row count. Then the pages are wrapped into independent transactions to achieve flexible paging updates of a book is not required on the book lists page but only required on the book details and order pages. Stale Read can be used to improve throughout of the application. 4.7.8.2.2 Statement level To query

compatibility with MySQL in the areas of security and pes- simistic transaction mode, refer to the dedicated pages on Security and Pes- simistic Transaction Mode. You can try out TiDB features on TiDB Playground unique index to sort results and the offset keyword in the LIMIT clause to split pages by a specified row count. Then the pages are wrapped into independent transactions to achieve flexible paging updates of a book is not required on the book lists page but only required on the book details and order pages. Stale Read can be used to improve throughout of the application. 4.7.8.2.2 Statement level To query