1====================================== 2NO_HZ: Reducing Scheduling-Clock Ticks 3====================================== 4 5 6This document describes Kconfig options and boot parameters that can 7reduce the number of scheduling-clock interrupts, thereby improving energy 8efficiency and reducing OS jitter. Reducing OS jitter is important for 9some types of computationally intensive high-performance computing (HPC) 10applications and for real-time applications. 11 12There are three main ways of managing scheduling-clock interrupts 13(also known as "scheduling-clock ticks" or simply "ticks"): 14 151. Never omit scheduling-clock ticks (CONFIG_HZ_PERIODIC=y or 16 CONFIG_NO_HZ=n for older kernels). You normally will -not- 17 want to choose this option. 18 192. Omit scheduling-clock ticks on idle CPUs (CONFIG_NO_HZ_IDLE=y or 20 CONFIG_NO_HZ=y for older kernels). This is the most common 21 approach, and should be the default. 22 233. Omit scheduling-clock ticks on CPUs that are either idle or that 24 have only one runnable task (CONFIG_NO_HZ_FULL=y). Unless you 25 are running realtime applications or certain types of HPC 26 workloads, you will normally -not- want this option. 27 28These three cases are described in the following three sections, followed 29by a third section on RCU-specific considerations, a fourth section 30discussing testing, and a fifth and final section listing known issues. 31 32 33Never Omit Scheduling-Clock Ticks 34================================= 35 36Very old versions of Linux from the 1990s and the very early 2000s 37are incapable of omitting scheduling-clock ticks. It turns out that 38there are some situations where this old-school approach is still the 39right approach, for example, in heavy workloads with lots of tasks 40that use short bursts of CPU, where there are very frequent idle 41periods, but where these idle periods are also quite short (tens or 42hundreds of microseconds). For these types of workloads, scheduling 43clock interrupts will normally be delivered any way because there 44will frequently be multiple runnable tasks per CPU. In these cases, 45attempting to turn off the scheduling clock interrupt will have no effect 46other than increasing the overhead of switching to and from idle and 47transitioning between user and kernel execution. 48 49This mode of operation can be selected using CONFIG_HZ_PERIODIC=y (or 50CONFIG_NO_HZ=n for older kernels). 51 52However, if you are instead running a light workload with long idle 53periods, failing to omit scheduling-clock interrupts will result in 54excessive power consumption. This is especially bad on battery-powered 55devices, where it results in extremely short battery lifetimes. If you 56are running light workloads, you should therefore read the following 57section. 58 59In addition, if you are running either a real-time workload or an HPC 60workload with short iterations, the scheduling-clock interrupts can 61degrade your applications performance. If this describes your workload, 62you should read the following two sections. 63 64 65Omit Scheduling-Clock Ticks For Idle CPUs 66========================================= 67 68If a CPU is idle, there is little point in sending it a scheduling-clock 69interrupt. After all, the primary purpose of a scheduling-clock interrupt 70is to force a busy CPU to shift its attention among multiple duties, 71and an idle CPU has no duties to shift its attention among. 72 73An idle CPU that is not receiving scheduling-clock interrupts is said to 74be "dyntick-idle", "in dyntick-idle mode", "in nohz mode", or "running 75tickless". The remainder of this document will use "dyntick-idle mode". 76 77The CONFIG_NO_HZ_IDLE=y Kconfig option causes the kernel to avoid sending 78scheduling-clock interrupts to idle CPUs, which is critically important 79both to battery-powered devices and to highly virtualized mainframes. 80A battery-powered device running a CONFIG_HZ_PERIODIC=y kernel would 81drain its battery very quickly, easily 2-3 times as fast as would the 82same device running a CONFIG_NO_HZ_IDLE=y kernel. A mainframe running 831,500 OS instances might find that half of its CPU time was consumed by 84unnecessary scheduling-clock interrupts. In these situations, there 85is strong motivation to avoid sending scheduling-clock interrupts to 86idle CPUs. That said, dyntick-idle mode is not free: 87 881. It increases the number of instructions executed on the path 89 to and from the idle loop. 90 912. On many architectures, dyntick-idle mode also increases the 92 number of expensive clock-reprogramming operations. 93 94Therefore, systems with aggressive real-time response constraints often 95run CONFIG_HZ_PERIODIC=y kernels (or CONFIG_NO_HZ=n for older kernels) 96in order to avoid degrading from-idle transition latencies. 97 98There is also a boot parameter "nohz=" that can be used to disable 99dyntick-idle mode in CONFIG_NO_HZ_IDLE=y kernels by specifying "nohz=off". 100By default, CONFIG_NO_HZ_IDLE=y kernels boot with "nohz=on", enabling 101dyntick-idle mode. 102 103 104Omit Scheduling-Clock Ticks For CPUs With Only One Runnable Task 105================================================================ 106 107If a CPU has only one runnable task, there is little point in sending it 108a scheduling-clock interrupt because there is no other task to switch to. 109Note that omitting scheduling-clock ticks for CPUs with only one runnable 110task implies also omitting them for idle CPUs. 111 112The CONFIG_NO_HZ_FULL=y Kconfig option causes the kernel to avoid 113sending scheduling-clock interrupts to CPUs with a single runnable task, 114and such CPUs are said to be "adaptive-ticks CPUs". This is important 115for applications with aggressive real-time response constraints because 116it allows them to improve their worst-case response times by the maximum 117duration of a scheduling-clock interrupt. It is also important for 118computationally intensive short-iteration workloads: If any CPU is 119delayed during a given iteration, all the other CPUs will be forced to 120wait idle while the delayed CPU finishes. Thus, the delay is multiplied 121by one less than the number of CPUs. In these situations, there is 122again strong motivation to avoid sending scheduling-clock interrupts. 123 124By default, no CPU will be an adaptive-ticks CPU. The "nohz_full=" 125boot parameter specifies the adaptive-ticks CPUs. For example, 126"nohz_full=1,6-8" says that CPUs 1, 6, 7, and 8 are to be adaptive-ticks 127CPUs. Note that you are prohibited from marking all of the CPUs as 128adaptive-tick CPUs: At least one non-adaptive-tick CPU must remain 129online to handle timekeeping tasks in order to ensure that system 130calls like gettimeofday() returns accurate values on adaptive-tick CPUs. 131(This is not an issue for CONFIG_NO_HZ_IDLE=y because there are no running 132user processes to observe slight drifts in clock rate.) Therefore, the 133boot CPU is prohibited from entering adaptive-ticks mode. Specifying a 134"nohz_full=" mask that includes the boot CPU will result in a boot-time 135error message, and the boot CPU will be removed from the mask. Note that 136this means that your system must have at least two CPUs in order for 137CONFIG_NO_HZ_FULL=y to do anything for you. 138 139Finally, adaptive-ticks CPUs must have their RCU callbacks offloaded. 140This is covered in the "RCU IMPLICATIONS" section below. 141 142Normally, a CPU remains in adaptive-ticks mode as long as possible. 143In particular, transitioning to kernel mode does not automatically change 144the mode. Instead, the CPU will exit adaptive-ticks mode only if needed, 145for example, if that CPU enqueues an RCU callback. 146 147Just as with dyntick-idle mode, the benefits of adaptive-tick mode do 148not come for free: 149 1501. CONFIG_NO_HZ_FULL selects CONFIG_NO_HZ_COMMON, so you cannot run 151 adaptive ticks without also running dyntick idle. This dependency 152 extends down into the implementation, so that all of the costs 153 of CONFIG_NO_HZ_IDLE are also incurred by CONFIG_NO_HZ_FULL. 154 1552. The user/kernel transitions are slightly more expensive due 156 to the need to inform kernel subsystems (such as RCU) about 157 the change in mode. 158 1593. POSIX CPU timers prevent CPUs from entering adaptive-tick mode. 160 Real-time applications needing to take actions based on CPU time 161 consumption need to use other means of doing so. 162 1634. If there are more perf events pending than the hardware can 164 accommodate, they are normally round-robined so as to collect 165 all of them over time. Adaptive-tick mode may prevent this 166 round-robining from happening. This will likely be fixed by 167 preventing CPUs with large numbers of perf events pending from 168 entering adaptive-tick mode. 169 1705. Scheduler statistics for adaptive-tick CPUs may be computed 171 slightly differently than those for non-adaptive-tick CPUs. 172 This might in turn perturb load-balancing of real-time tasks. 173 174Although improvements are expected over time, adaptive ticks is quite 175useful for many types of real-time and compute-intensive applications. 176However, the drawbacks listed above mean that adaptive ticks should not 177(yet) be enabled by default. 178 179 180RCU Implications 181================ 182 183There are situations in which idle CPUs cannot be permitted to 184enter either dyntick-idle mode or adaptive-tick mode, the most 185common being when that CPU has RCU callbacks pending. 186 187Avoid this by offloading RCU callback processing to "rcuo" kthreads 188using the CONFIG_RCU_NOCB_CPU=y Kconfig option. The specific CPUs to 189offload may be selected using The "rcu_nocbs=" kernel boot parameter, 190which takes a comma-separated list of CPUs and CPU ranges, for example, 191"1,3-5" selects CPUs 1, 3, 4, and 5. Note that CPUs specified by 192the "nohz_full" kernel boot parameter are also offloaded. 193 194The offloaded CPUs will never queue RCU callbacks, and therefore RCU 195never prevents offloaded CPUs from entering either dyntick-idle mode 196or adaptive-tick mode. That said, note that it is up to userspace to 197pin the "rcuo" kthreads to specific CPUs if desired. Otherwise, the 198scheduler will decide where to run them, which might or might not be 199where you want them to run. 200 201 202Testing 203======= 204 205So you enable all the OS-jitter features described in this document, 206but do not see any change in your workload's behavior. Is this because 207your workload isn't affected that much by OS jitter, or is it because 208something else is in the way? This section helps answer this question 209by providing a simple OS-jitter test suite, which is available on branch 210master of the following git archive: 211 212git://git.kernel.org/pub/scm/linux/kernel/git/frederic/dynticks-testing.git 213 214Clone this archive and follow the instructions in the README file. 215This test procedure will produce a trace that will allow you to evaluate 216whether or not you have succeeded in removing OS jitter from your system. 217If this trace shows that you have removed OS jitter as much as is 218possible, then you can conclude that your workload is not all that 219sensitive to OS jitter. 220 221Note: this test requires that your system have at least two CPUs. 222We do not currently have a good way to remove OS jitter from single-CPU 223systems. 224 225 226Known Issues 227============ 228 229* Dyntick-idle slows transitions to and from idle slightly. 230 In practice, this has not been a problem except for the most 231 aggressive real-time workloads, which have the option of disabling 232 dyntick-idle mode, an option that most of them take. However, 233 some workloads will no doubt want to use adaptive ticks to 234 eliminate scheduling-clock interrupt latencies. Here are some 235 options for these workloads: 236 237 a. Use PMQOS from userspace to inform the kernel of your 238 latency requirements (preferred). 239 240 b. On x86 systems, use the "idle=mwait" boot parameter. 241 242 c. On x86 systems, use the "intel_idle.max_cstate=" to limit 243 ` the maximum C-state depth. 244 245 d. On x86 systems, use the "idle=poll" boot parameter. 246 However, please note that use of this parameter can cause 247 your CPU to overheat, which may cause thermal throttling 248 to degrade your latencies -- and that this degradation can 249 be even worse than that of dyntick-idle. Furthermore, 250 this parameter effectively disables Turbo Mode on Intel 251 CPUs, which can significantly reduce maximum performance. 252 253* Adaptive-ticks slows user/kernel transitions slightly. 254 This is not expected to be a problem for computationally intensive 255 workloads, which have few such transitions. Careful benchmarking 256 will be required to determine whether or not other workloads 257 are significantly affected by this effect. 258 259* Adaptive-ticks does not do anything unless there is only one 260 runnable task for a given CPU, even though there are a number 261 of other situations where the scheduling-clock tick is not 262 needed. To give but one example, consider a CPU that has one 263 runnable high-priority SCHED_FIFO task and an arbitrary number 264 of low-priority SCHED_OTHER tasks. In this case, the CPU is 265 required to run the SCHED_FIFO task until it either blocks or 266 some other higher-priority task awakens on (or is assigned to) 267 this CPU, so there is no point in sending a scheduling-clock 268 interrupt to this CPU. However, the current implementation 269 nevertheless sends scheduling-clock interrupts to CPUs having a 270 single runnable SCHED_FIFO task and multiple runnable SCHED_OTHER 271 tasks, even though these interrupts are unnecessary. 272 273 And even when there are multiple runnable tasks on a given CPU, 274 there is little point in interrupting that CPU until the current 275 running task's timeslice expires, which is almost always way 276 longer than the time of the next scheduling-clock interrupt. 277 278 Better handling of these sorts of situations is future work. 279 280* A reboot is required to reconfigure both adaptive idle and RCU 281 callback offloading. Runtime reconfiguration could be provided 282 if needed, however, due to the complexity of reconfiguring RCU at 283 runtime, there would need to be an earthshakingly good reason. 284 Especially given that you have the straightforward option of 285 simply offloading RCU callbacks from all CPUs and pinning them 286 where you want them whenever you want them pinned. 287 288* Additional configuration is required to deal with other sources 289 of OS jitter, including interrupts and system-utility tasks 290 and processes. This configuration normally involves binding 291 interrupts and tasks to particular CPUs. 292 293* Some sources of OS jitter can currently be eliminated only by 294 constraining the workload. For example, the only way to eliminate 295 OS jitter due to global TLB shootdowns is to avoid the unmapping 296 operations (such as kernel module unload operations) that 297 result in these shootdowns. For another example, page faults 298 and TLB misses can be reduced (and in some cases eliminated) by 299 using huge pages and by constraining the amount of memory used 300 by the application. Pre-faulting the working set can also be 301 helpful, especially when combined with the mlock() and mlockall() 302 system calls. 303 304* Unless all CPUs are idle, at least one CPU must keep the 305 scheduling-clock interrupt going in order to support accurate 306 timekeeping. 307 308* If there might potentially be some adaptive-ticks CPUs, there 309 will be at least one CPU keeping the scheduling-clock interrupt 310 going, even if all CPUs are otherwise idle. 311 312 Better handling of this situation is ongoing work. 313 314* Some process-handling operations still require the occasional 315 scheduling-clock tick. These operations include calculating CPU 316 load, maintaining sched average, computing CFS entity vruntime, 317 computing avenrun, and carrying out load balancing. They are 318 currently accommodated by scheduling-clock tick every second 319 or so. On-going work will eliminate the need even for these 320 infrequent scheduling-clock ticks. 321