@@ -595,7 +595,7 @@ coscheduling 协同调度是为了解决云服务场景, 为不同用户提供
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|:----:|:----:|:---:|:----:|:---------:|:----:|
| 2018/09/07 | Jan H. Schönherr | [[RFC,00/60] Coscheduling for Linux](https://lore.kernel.org/lkml/20180907214047.26914-1-jschoenh@amazon .de) | 亚马逊的协同调度方案 | RFC ☐ | [LORE RFC,00/60 ](https://lore.kernel.org/lkml/20180907214047.26914-1-jschoenh@amazon.de ) |
| 2018/09/07 | Jan H. Schönherr | [[RFC,00/60] Coscheduling for Linux](https://lore.kernel.org/lkml/20180907214047.26914-1-jschoenh@amazon .de) | 亚马逊的协同调度方案. 参见 LWN 报道 [ Coscheduling: simultaneous scheduling in control groups ]( https://lwn.net/Articles/764482 ) 以及 [ Linux Coscheduling调度器简介 ]( https://blog.csdn.net/dog250/article/details/101752290 ). | RFC ☐ | [LORE RFC,00/60 ](https://lore.kernel.org/lkml/20180907214047.26914-1-jschoenh@amazon.de ) |
| 2019/02/18 | Peter & Kirill Tkhai 等 | [Core scheduling ](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/?id=67692435c411e5c53a1c588ecca2037aebd81f2e ) | 核调度器, 限制同一个 SMT 域内的两个 CPU 只能运行同一组进程.<br>1. v3 合入了 8 个重构的补丁.<br>2. | v9 ☑ 5.4-rc1 | [2019/02/18 LORE RFC ](https://lore.kernel.org/lkml/20190218165620.383905466@infradead.org )<br>*-*-*-*-*-*-*-* <br>[2019/04/23 LORE v2,00/17 ](https://lore.kernel.org/lkml/cover.1556025155.git.vpillai@digitalocean.com )<br>*-*-*-*-*-*-*-* <br>[2019/05/29 LORE v3,00/16 ](https://lore.kernel.org/all/cover.1559129225.git.vpillai@digitalocean.com )<br>*-*-*-*-*-*-*-* <br>[2019/10/30 LORE v4,00/19 ](https://lore.kernel.org/lkml/cover.1572437285.git.vpillai@digitalocean.com )<br>*-*-*-*-*-*-*-* <br>[2020/03/04 LORE v5,00/13 ](https://lore.kernel.org/lkml/cover.1583332764.git.vpillai@digitalocean.com )<br>*-*-*-*-*-*-*-* <br>[2020/06/30 LORE v6,00/16 ](https://lore.kernel.org/lkml/cover.1593530334.git.vpillai@digitalocean.com )<br>*-*-*-*-*-*-*-* <br>[2020/08/28 LORE v7,00/23 ](https://lore.kernel.org/lkml/cover.1598643276.git.jdesfossez@digitalocean.com )<br>*-*-*-*-*-*-*-* <br>[2020/10/19 LORE v8,00/26 ](https://lore.kernel.org/lkml/20201020014336.2076526-1-joel@joelfernandes.org )<br>*-*-*-*-*-*-*-* <br>[2020/11/17 LORE v9,00/32 ](https://lore.kernel.org/lkml/20201117232003.3580179-1-joel@joelfernandes.org ) |
| 2021/03/25 | Joel Fernandes 等 | [Core scheduling remaining patches rebase ](https://lore.kernel.org/patchwork/cover/1369931 ) | Core scheduling v9 的大部分补丁都已经在 TIP 分支了, 部分未合入补丁的重构与适配. | v10 ☐ | [2021/01/23 LORE v10,0/5 ](https://lore.kernel.org/lkml/20210123011704.1901835-1-joel@joelfernandes.org )<br>*-*-*-*-*-*-*-* <br>[2021/03/24 LORE v10 resend,0/8 ](https://lore.kernel.org/lkml/20210324214020.34142-1-joel@joelfernandes.org ) |
| 2021/04/01 | Peter Zijlstra | [sched: Core scheduling interfaces ](https://lore.kernel.org/lkml/20210401131012.395311786@infradead.org/ ) | Peter 将 TIP 分支 core scheduling 进行了重构. 重新设计了 Core scheduling 的接口. | v11 ☐ |[2021/04/01 v11,0/9 ](https://lore.kernel.org/lkml/20210401131012.395311786@infradead.org ) |
@@ -3797,8 +3797,8 @@ Oracle 数据库具有类似的虚拟化功能, 称为 Oracle Multitenant, 其
调度器最核心的工作就两个内容 :
- [x] * * 选进程 * * : 选择下一个更合适的进程 ** pick_next_task**
- [x] * * 选核 * * : 为某个进程选择更合适的 CPU 运行 ** select_task_rq**
- [x] * * 选进程 * * : 选择下一个更合适的进程 pick_next_task/Pick Next/Task Selection
- [x] * * 选核 * * : 为某个进程选择更合适的 CPU 运行 select_task_rq/Task Placement
所有其他的机制都是直接或者间接服务这两个终极任务的.
@@ -4015,6 +4015,7 @@ y = (1 - \frac{pct^{2}}{10000^{2}} \times x^{2}) \times llc\_weight
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|:----:|:----:|:----:|:---:|:----------:|:---:|
| 2021/12/01 | Mel Gorman <mgorman@techsingularity .net> | [Adjust NUMA imbalance for multiple LLCs ](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/?id=e496132ebedd870b67f1f6d2428f9bb9d7ae27fd ) | [commit 7d2b5dd0bcc4 ("sched/numa: Allow a floating imbalance between NUMA nodes") ](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7d2b5dd0bcc4 ) 允许 NUMA 节点之间的不平衡, 这样通信任务不会被 load balance 分开. 当 LLC 和 node 之间有 1:1 的关系时, 这种方法可以很好地工作, 但是对于多个 LLC, 如果独立的任务过早地使用 CPU 共享缓存, 这种方法就不太理想了. 本系列解决了两个问题:<br>1. 调度程序域权重的使用不一致, 以及当每个 NUMA 节点有许多 LLC 时性能不佳. NUMA 之间允许的不均衡的进程数目不再是一个固定的值 NUMA_IMBALANCE_MIN(2), 而是在 build_sched_domains() 中实际探测 NUMA 域下辖的 LLC 的数目, 作为 sd->imb_numa_nr. | v4 ☑✓ 5.18-rc1 | [PatchWork v3,0/2 ](https://lore.kernel.org/lkml/20211201151844.20488-1-mgorman@techsingularity.net )<br>*-*-*-*-*-*-*-* <br>[LORE v4,0/2 ](https://lore.kernel.org/lkml/20211210093307.31701-1-mgorman@techsingularity.net )<br>*-*-*-*-*-*-*-* <br>[LORE v6,0/2 ](https://lore.kernel.org/all/20220208094334.16379-1-mgorman@techsingularity.net ) |
| 2023/05/30 | Peter Zijlstra <peterz@infradead .org> | [sched/fair: Multi-LLC select_idle_sibling() ](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c7dfd6b9122d29d0e9a4587ab470c0564d7f92ab ) | Tejun 报告: Zen2 机器上有 4 LLC, 每个 LLC 上 3 个 core 的机器时, 发现系统很繁忙的情况下仍然出现大量 CPU 空间.<br>分析发现, 这是因为 select_idle_sibling() 不会考虑本地 LLC 之外的任何事情, 而且由于所有这些任务都是短时间运行的, 所以周期性空闲负载平衡器失效了.<br>虽然将工作缓存保存在本地很好, 但最好不要有大量的空闲时间. 因此, 当本地的 LLC 为空时, 让 select_idle_sibling() 尝试在同一节点内的其他 LLC. 参见 phoronix 的报道 [Linux 6.5 Scheduler Patch Will To Help AMD Systems With Multiple LLCs Per Die ](https://www.phoronix.com/news/Multi-LLC-Select-Idle-Sibling ). | v1 ☐☑✓ | [COMMIT ](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c7dfd6b9122d29d0e9a4587ab470c0564d7f92ab ) |
## 5.6 相关学术研究
@@ -5707,7 +5708,7 @@ PREEMPT-RT PATCH 的核心思想是最小化内核中不可抢占部分的代码
之前已经使用了许多方法来尝试改进重要进程的响应时间, 例如, 可以使用传统的 nice 值来提高进程的优先级, 这是可行的, 但是进程的友好性并不直接转化为延迟; 它控制进程可用的 CPU 时间的消耗, 但不能控制进程的调度延迟. 使用实时优先级将使调度程序快速运行进程, 解决进程的时延诉求. 但是也存在诸多问题.
### 8.9.1 l atency_n ice
### 8.9.1 L atency N ice & EEVDF
-------
| 日期 | LWN | 翻译 |
@@ -5717,7 +5718,11 @@ PREEMPT-RT PATCH 的核心思想是最小化内核中不可抢占部分的代码
| 2022/03/17 | [Improved response times with latency nice ](https://lwn.net/Articles/887842 ) | [LWN: 采用 latency nice 改善响应时间 ](https://blog.csdn.net/Linux_Everything/article/details/123887454 ) |
| 2022/04/05 | NA | 国内对这组补丁的分析 [latency-nice 优先级补丁源码分析 ](https://blog.csdn.net/qq_23662505/article/details/123977540 ) |
| 2022/09/13 | LPC-2022 上关于 latency nice 的演讲: [Latency hints for CFS task ](https://lpc.events/event/16/contributions/1273 ) | NA |
| 2023/03/09 | [An EEVDF CPU scheduler for Linux ](https://lwn.net/Articles/925371 ) |
| 2023/03/09 | [An EEVDF CPU scheduler for Linux ](https://lwn.net/Articles/925371 ), 论文 [ Earliest Eligible Virtual Deadline First A Flexible ]( https://people.eecs.berkeley.edu/~istoica/papers/eevdf-tr-95.pdf ), [ 翻译-LWN： ]( https://blog.csdn.net/Linux_Everything/article/details/129807209 ), [ Updated EEVDF Linux CPU Scheduler Patches Posted That Plan To Replace CFS ]( https://www.phoronix.com/news/EEVDF-Scheduler-Linux-EO-May ) |
#### 8.9.1.1 Latency Nice
-------
2020 年, Parth Shah 提出了 latency nice 的概念. 旨在对应用的延迟进行感知和标记, 降低延迟敏感应用程序的调度延迟, 使其更快地获得 CPU 时间. latency_nice 值与现有 nice 值相对应, 介于 -20 和 19 之间. 数字越低, 优先级越高.
@@ -5768,8 +5773,17 @@ enqueue_task_fair()
|:-----:|:----:|:----:|:----:|:------------:|:----:|
| 2020/02/28 | Parth Shah <parth@linux .ibm.com> | [Introduce per-task latency_nice for scheduler hints ](https://lore.kernel.org/all/20200228090755.22829-1-parth@linux.ibm.com ) | 20200228090755.22829-1-parth@linux .ibm.com | v5 ☐☑✓ | [LORE v4,0/4 ](https://lore.kernel.org/lkml/20200224085918.16955-1-parth@linux.ibm.com )<br>*-*-*-*-*-*-*-* <br>[LORE v5,0/4 ](https://lore.kernel.org/all/20200228090755.22829-1-parth@linux.ibm.com ) |
| 2020/05/07 | Parth Shah <parth@linux .ibm.com> | [IDLE gating in presence of latency-sensitive tasks ](https://lore.kernel.org/all/20200507133723.18325-1-parth@linux.ibm.com ) | 20200507133723.18325-1-parth@linux .ibm.com | v1 ☐☑✓ | [LORE v1,0/4 ](https://lore.kernel.org/all/20200507133723.18325-1-parth@linux.ibm.com ) |
| 2022/09/16 | Vincent Guittot <vincent.guittot@linaro .org> | [Add latency_nice priority ](https://lore.kernel.org/all/20220311161406.23497-1-vincent.guittot@linaro.org ) | 参见 [Improved response times with latency nice ](https://lwn.net/Articles/887842 ). | v1 ☐☑✓ | [2022/03/11 LORE v1,0/6 ](https://lore.kernel.org/all/20220311161406.23497-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/05/12 LORE v2,0/7 ](https://lore.kernel.org/all/20220512163534.2572-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/09/09 LORE v3,0/8 ](https://lore.kernel.org/all/20220909130309.25458-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/09/16 LORE v4,0/8 ](https://lore.kernel.org/all/20220916080305.29574-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/09/25 LORE v5,0/8 ](https://lore.kernel.org/all/20220925143908.10846-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/10/28 LORE v6,0/9 ](https://lore.kernel.org/all/20221028073637.31195-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/10/28 LORE v7,0/9 ](https://lore.kernel.org/all/20221028093403.6673-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/11/10 LORE v8,0/9 ](https://lore.kernel.org/all/20221110175009.18458-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/11/15 LORE v9,0/9 ](https://lore.kernel.org/all/20221115171851.835-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2023/01/13 LORE v10,0/9 ](https://lore.kernel.org/all/20230113141234.260128-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2023/02/23 LORE v11,0/9 ](https://lore.kernel.org/all/20230223191041.577305-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2023/02/24 LORE v12,0/8 ](https://lore.kernel.org/all/20230224093454.956298-1-vincent.guittot@linaro.org/ ) |
| 2022/09/16 | Vincent Guittot <vincent.guittot@linaro .org> | [Add latency_nice priority ](https://lore.kernel.org/all/20220311161406.23497-1-vincent.guittot@linaro.org ) | 参见 [Improved response times with latency nice ](https://lwn.net/Articles/887842 ). | v1 ☐☑✓ | [2022/03/11 LORE v1,0/6 ](https://lore.kernel.org/all/20220311161406.23497-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/05/12 LORE v2,0/7 ](https://lore.kernel.org/all/20220512163534.2572-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/09/09 LORE v3,0/8 ](https://lore.kernel.org/all/20220909130309.25458-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/09/16 LORE v4,0/8 ](https://lore.kernel.org/all/20220916080305.29574-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/09/25 LORE v5,0/8 ](https://lore.kernel.org/all/20220925143908.10846-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/10/28 LORE v6,0/9 ](https://lore.kernel.org/all/20221028073637.31195-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/10/28 LORE v7,0/9 ](https://lore.kernel.org/all/20221028093403.6673-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/11/10 LORE v8,0/9 ](https://lore.kernel.org/all/20221110175009.18458-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2022/11/15 LORE v9,0/9 ](https://lore.kernel.org/all/20221115171851.835-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2023/01/13 LORE v10,0/9 ](https://lore.kernel.org/all/20230113141234.260128-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2023/02/23 LORE v11,0/9 ](https://lore.kernel.org/all/20230223191041.577305-1-vincent.guittot@linaro.org )<br>*-*-*-*-*-*-*-* <br>[2023/02/24 LORE v12,0/8 ](https://lore.kernel.org/all/20230224093454.956298-1-vincent.guittot@linaro.org ) |
#### 8.9.1.2 EEVDF
-------
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|:-----:|:----:|:----:|:----:|:------------:|:----:|
| 2023/03/28 | Peter Zijlstra <peterz@infradead .org> | [sched: EEVDF using latency-nice ](https://lore.kernel.org/all/20230328092622.062917921@infradead.org ) | TODO | v1 ☐☑✓ | [LORE 00/10 ](https://lore.kernel.org/all/20230306132521.968182689@infradead.org )<br>*-*-*-*-*-*-*-* <br>[LORE v1,0/17 ](https://lore.kernel.org/all/20230328092622.062917921@infradead.org ) |
| 2023/05/31 | Peter Zijlstra <peterz@infradead .org> | [sched: EEVDF and latency-nice and/or slice-attr ](https://lore.kernel.org/all/20230531115839.089944915@infradead.org ) | TODO | v1 ☐☑✓ | [LORE v1,0/15 ](https://lore.kernel.org/all/20230531115839.089944915@infradead.org ) |
### 8.9.2 Xen CPU Scheduling
@@ -6255,6 +6269,10 @@ B 站 Plugsched 介绍视频 [纯干货解读：Plugsched, 首次实现 Linux ke
| [CacULE CPU Scheduler ](https://github.com/hamadmarri/cacule-cpu-scheduler ) | 是基于交互性评分机制 (CacULE Interactivity Score) 的 CFS 补丁集. 交互性分数的灵感来自 FreeBSD 的 ULE 调度器, 可以增强系统响应能力 / 延迟. ARCHLINUX 的开发者 ptr1337, 同样移植了 [CacULE Scheduler ](https://github.com/ptr1337/linux-cacule ), [CachyOS/linux-cachyos ](https://github.com/CachyOS/linux-cachyos ). 使用 calc_interactivity(), calc_cache_score(), calc_starve_score() 分别计算任务的交互性评分 (Interactivity Score), 缓存亲和性评分(Cache Score) 以及饥饿评分(Starve Score), 然后将 CFS 的选取下一个任务机制 pick_next_task_fair 更改为 ULE 的评分机制, 以便选取要运行的下一个任务. |
| [Task Type(TT) CPU Scheduler ](https://github.com/hamadmarri/TT-CPU-Scheduler ) | 根据任务的行为检测并识别任务类型, 并根据其类型控制调度. 基于任务类型的好处是允许调度程序进行更多控制, 并选择接下来在 CPU 中运行的最佳任务. 当前有 5 种类型: 实时(REALTIME), 交互(INTERACTIVE), 无类型(NO_TYPE), 计算密集型(CPU_BOUND), 批处理(BATCH). 调度器通过 detect_type() 周期性地探测应用的 task_type. |
| [Baby-CPU-Scheduler ](https://github.com/hamadmarri/Baby-CPU-Scheduler ) | 一个非常基本, 轻量级但性能非常高的调度器 Basic Scheduler (BS). 可以将其用作 Linux 上的基本调度程序进行学习 |
| [Burst-Oriented Response Enhancer (BORE) CPU Scheduler ](https://github.com/firelzrd/bore-scheduler ) | BORE(面向突发的响应增强器)是 CFS(完全公平调度程序)的增强版本, CFS 是 Linux 中默认的 CPU 调度程序, 旨在保持 CFS 的高吞吐量性能, 同时在尽可能宽的负载情况下提供对用户输入的更高响应能力. 为了实现这一目标, BORE 为每个单独的任务引入了一个称为 "突发性" 的灵活性维度, 部分偏离了 CFS 固有的 "完全公平" 原则. 延迟是指任务在显式放弃后通过进入睡眠、IO 等待或屈服而消耗的累积 CPU 时间得出的分数. 该分数代表了广泛的时间特征, 从纳秒到数百秒不等, 在不同的任务中有所不同. |
| [Multiple run-queues for BFS ](https://lore.kernel.org/all/1355591803.23863.3.camel@findus-T530 ) | [两个非常有意思的适合桌面使用的Linux task调度器: BFS和MuqSS ](https://blog.csdn.net/juS3Ve/article/details/102380529 )<br>[操作系统调度算法5——MuQss，多队列跳表调度器 ](https://zhuanlan.zhihu.com/p/373693657 ), [ckolivas/linux ](https://github.com/ckolivas/linux ), [CK 的博客 ](http://ck-hack.blogspot.com ). [细说CFS与MuQSS以及load-balance ](https://blog.csdn.net/qq_23662505/article/details/120220689 ), [MuQSS_调度器-The_MuQSS_CPU_scheduler ](https://github.com/jiebaomaster/linux-kernel-doc-translate/blob/master/lwn/MuQSS_调度器-The_MuQSS_CPU_scheduler.md ) |
#### 11.2.5.2 User Sensitive Factors for Scheduler
-------
Cheng Jian