Documentation/timers/NO_HZ.txt

   1                 NO_HZ: Reducing Scheduling-Clock Ticks
   2
   3
   4 This document describes Kconfig options and boot parameters that can
   5 reduce the number of scheduling-clock interrupts, thereby improving energy
   6 efficiency and reducing OS jitter.  Reducing OS jitter is important for
   7 some types of computationally intensive high-performance computing (HPC)
   8 applications and for real-time applications.
   9
  10 There are three main ways of managing scheduling-clock interrupts
  11 (also known as "scheduling-clock ticks" or simply "ticks"):
  12
  13 1.      Never omit scheduling-clock ticks (CONFIG_HZ_PERIODIC=y or
  14         CONFIG_NO_HZ=n for older kernels).  You normally will -not-
  15         want to choose this option.
  16
  17 2.      Omit scheduling-clock ticks on idle CPUs (CONFIG_NO_HZ_IDLE=y or
  18         CONFIG_NO_HZ=y for older kernels).  This is the most common
  19         approach, and should be the default.
  20
  21 3.      Omit scheduling-clock ticks on CPUs that are either idle or that
  22         have only one runnable task (CONFIG_NO_HZ_FULL=y).  Unless you
  23         are running realtime applications or certain types of HPC
  24         workloads, you will normally -not- want this option.
  25
  26 These three cases are described in the following three sections, followed
  27 by a third section on RCU-specific considerations and a fourth and final
  28 section listing known issues.
  29
  30
  31 NEVER OMIT SCHEDULING-CLOCK TICKS
  32
  33 Very old versions of Linux from the 1990s and the very early 2000s
  34 are incapable of omitting scheduling-clock ticks.  It turns out that
  35 there are some situations where this old-school approach is still the
  36 right approach, for example, in heavy workloads with lots of tasks
  37 that use short bursts of CPU, where there are very frequent idle
  38 periods, but where these idle periods are also quite short (tens or
  39 hundreds of microseconds).  For these types of workloads, scheduling
  40 clock interrupts will normally be delivered any way because there
  41 will frequently be multiple runnable tasks per CPU.  In these cases,
  42 attempting to turn off the scheduling clock interrupt will have no effect
  43 other than increasing the overhead of switching to and from idle and
  44 transitioning between user and kernel execution.
  45
  46 This mode of operation can be selected using CONFIG_HZ_PERIODIC=y (or
  47 CONFIG_NO_HZ=n for older kernels).
  48
  49 However, if you are instead running a light workload with long idle
  50 periods, failing to omit scheduling-clock interrupts will result in
  51 excessive power consumption.  This is especially bad on battery-powered
  52 devices, where it results in extremely short battery lifetimes.  If you
  53 are running light workloads, you should therefore read the following
  54 section.
  55
  56 In addition, if you are running either a real-time workload or an HPC
  57 workload with short iterations, the scheduling-clock interrupts can
  58 degrade your applications performance.  If this describes your workload,
  59 you should read the following two sections.
  60
  61
  62 OMIT SCHEDULING-CLOCK TICKS FOR IDLE CPUs
  63
  64 If a CPU is idle, there is little point in sending it a scheduling-clock
  65 interrupt.  After all, the primary purpose of a scheduling-clock interrupt
  66 is to force a busy CPU to shift its attention among multiple duties,
  67 and an idle CPU has no duties to shift its attention among.
  68
  69 The CONFIG_NO_HZ_IDLE=y Kconfig option causes the kernel to avoid sending
  70 scheduling-clock interrupts to idle CPUs, which is critically important
  71 both to battery-powered devices and to highly virtualized mainframes.
  72 A battery-powered device running a CONFIG_HZ_PERIODIC=y kernel would
  73 drain its battery very quickly, easily 2-3 times as fast as would the
  74 same device running a CONFIG_NO_HZ_IDLE=y kernel.  A mainframe running
  75 1,500 OS instances might find that half of its CPU time was consumed by
  76 unnecessary scheduling-clock interrupts.  In these situations, there
  77 is strong motivation to avoid sending scheduling-clock interrupts to
  78 idle CPUs.  That said, dyntick-idle mode is not free:
  79
  80 1.      It increases the number of instructions executed on the path
  81         to and from the idle loop.
  82
  83 2.      On many architectures, dyntick-idle mode also increases the
  84         number of expensive clock-reprogramming operations.
  85
  86 Therefore, systems with aggressive real-time response constraints often
  87 run CONFIG_HZ_PERIODIC=y kernels (or CONFIG_NO_HZ=n for older kernels)
  88 in order to avoid degrading from-idle transition latencies.
  89
  90 An idle CPU that is not receiving scheduling-clock interrupts is said to
  91 be "dyntick-idle", "in dyntick-idle mode", "in nohz mode", or "running
  92 tickless".  The remainder of this document will use "dyntick-idle mode".
  93
  94 There is also a boot parameter "nohz=" that can be used to disable
  95 dyntick-idle mode in CONFIG_NO_HZ_IDLE=y kernels by specifying "nohz=off".
  96 By default, CONFIG_NO_HZ_IDLE=y kernels boot with "nohz=on", enabling
  97 dyntick-idle mode.
  98
  99
 100 OMIT SCHEDULING-CLOCK TICKS FOR CPUs WITH ONLY ONE RUNNABLE TASK
 101
 102 If a CPU has only one runnable task, there is little point in sending it
 103 a scheduling-clock interrupt because there is no other task to switch to.
 104 Note that omitting scheduling-clock ticks for CPUs with only one runnable
 105 task implies also omitting them for idle CPUs.
 106
 107 The CONFIG_NO_HZ_FULL=y Kconfig option causes the kernel to avoid
 108 sending scheduling-clock interrupts to CPUs with a single runnable task,
 109 and such CPUs are said to be "adaptive-ticks CPUs".  This is important
 110 for applications with aggressive real-time response constraints because
 111 it allows them to improve their worst-case response times by the maximum
 112 duration of a scheduling-clock interrupt.  It is also important for
 113 computationally intensive short-iteration workloads:  If any CPU is
 114 delayed during a given iteration, all the other CPUs will be forced to
 115 wait idle while the delayed CPU finishes.  Thus, the delay is multiplied
 116 by one less than the number of CPUs.  In these situations, there is
 117 again strong motivation to avoid sending scheduling-clock interrupts.
 118
 119 By default, no CPU will be an adaptive-ticks CPU.  The "nohz_full="
 120 boot parameter specifies the adaptive-ticks CPUs.  For example,
 121 "nohz_full=1,6-8" says that CPUs 1, 6, 7, and 8 are to be adaptive-ticks
 122 CPUs.  Note that you are prohibited from marking all of the CPUs as
 123 adaptive-tick CPUs:  At least one non-adaptive-tick CPU must remain
 124 online to handle timekeeping tasks in order to ensure that system calls
 125 like gettimeofday() returns accurate values on adaptive-tick CPUs.
 126 (This is not an issue for CONFIG_NO_HZ_IDLE=y because there are no
 127 running user processes to observe slight drifts in clock rate.)
 128 Therefore, the boot CPU is prohibited from entering adaptive-ticks
 129 mode.  Specifying a "nohz_full=" mask that includes the boot CPU will
 130 result in a boot-time error message, and the boot CPU will be removed
 131 from the mask.
 132
 133 Alternatively, the CONFIG_NO_HZ_FULL_ALL=y Kconfig parameter specifies
 134 that all CPUs other than the boot CPU are adaptive-ticks CPUs.  This
 135 Kconfig parameter will be overridden by the "nohz_full=" boot parameter,
 136 so that if both the CONFIG_NO_HZ_FULL_ALL=y Kconfig parameter and
 137 the "nohz_full=1" boot parameter is specified, the boot parameter will
 138 prevail so that only CPU 1 will be an adaptive-ticks CPU.
 139
 140 Finally, adaptive-ticks CPUs must have their RCU callbacks offloaded.
 141 This is covered in the "RCU IMPLICATIONS" section below.
 142
 143 Normally, a CPU remains in adaptive-ticks mode as long as possible.
 144 In particular, transitioning to kernel mode does not automatically change
 145 the mode.  Instead, the CPU will exit adaptive-ticks mode only if needed,
 146 for example, if that CPU enqueues an RCU callback.
 147
 148 Just as with dyntick-idle mode, the benefits of adaptive-tick mode do
 149 not come for free:
 150
 151 1.      CONFIG_NO_HZ_FULL selects CONFIG_NO_HZ_COMMON, so you cannot run
 152         adaptive ticks without also running dyntick idle.  This dependency
 153         extends down into the implementation, so that all of the costs
 154         of CONFIG_NO_HZ_IDLE are also incurred by CONFIG_NO_HZ_FULL.
 155
 156 2.      The user/kernel transitions are slightly more expensive due
 157         to the need to inform kernel subsystems (such as RCU) about
 158         the change in mode.
 159
 160 3.      POSIX CPU timers on adaptive-tick CPUs may miss their deadlines
 161         (perhaps indefinitely) because they currently rely on
 162         scheduling-tick interrupts.  This will likely be fixed in
 163         one of two ways: (1) Prevent CPUs with POSIX CPU timers from
 164         entering adaptive-tick mode, or (2) Use hrtimers or other
 165         adaptive-ticks-immune mechanism to cause the POSIX CPU timer to
 166         fire properly.
 167
 168 4.      If there are more perf events pending than the hardware can
 169         accommodate, they are normally round-robined so as to collect
 170         all of them over time.  Adaptive-tick mode may prevent this
 171         round-robining from happening.  This will likely be fixed by
 172         preventing CPUs with large numbers of perf events pending from
 173         entering adaptive-tick mode.
 174
 175 5.      Scheduler statistics for adaptive-tick CPUs may be computed
 176         slightly differently than those for non-adaptive-tick CPUs.
 177         This might in turn perturb load-balancing of real-time tasks.
 178
 179 6.      The LB_BIAS scheduler feature is disabled by adaptive ticks.
 180
 181 Although improvements are expected over time, adaptive ticks is quite
 182 useful for many types of real-time and compute-intensive applications.
 183 However, the drawbacks listed above mean that adaptive ticks should not
 184 (yet) be enabled by default.
 185
 186
 187 RCU IMPLICATIONS
 188
 189 There are situations in which idle CPUs cannot be permitted to
 190 enter either dyntick-idle mode or adaptive-tick mode, the most
 191 common being when that CPU has RCU callbacks pending.
 192
 193 The CONFIG_RCU_FAST_NO_HZ=y Kconfig option may be used to cause such CPUs
 194 to enter dyntick-idle mode or adaptive-tick mode anyway.  In this case,
 195 a timer will awaken these CPUs every four jiffies in order to ensure
 196 that the RCU callbacks are processed in a timely fashion.
 197
 198 Another approach is to offload RCU callback processing to "rcuo" kthreads
 199 using the CONFIG_RCU_NOCB_CPU=y Kconfig option.  The specific CPUs to
 200 offload may be selected via several methods:
 201
 202 1.      One of three mutually exclusive Kconfig options specify a
 203         build-time default for the CPUs to offload:
 204
 205         a.      The CONFIG_RCU_NOCB_CPU_NONE=y Kconfig option results in
 206                 no CPUs being offloaded.
 207
 208         b.      The CONFIG_RCU_NOCB_CPU_ZERO=y Kconfig option causes
 209                 CPU 0 to be offloaded.
 210
 211         c.      The CONFIG_RCU_NOCB_CPU_ALL=y Kconfig option causes all
 212                 CPUs to be offloaded.  Note that the callbacks will be
 213                 offloaded to "rcuo" kthreads, and that those kthreads
 214                 will in fact run on some CPU.  However, this approach
 215                 gives fine-grained control on exactly which CPUs the
 216                 callbacks run on, along with their scheduling priority
 217                 (including the default of SCHED_OTHER), and it further
 218                 allows this control to be varied dynamically at runtime.
 219
 220 2.      The "rcu_nocbs=" kernel boot parameter, which takes a comma-separated
 221         list of CPUs and CPU ranges, for example, "1,3-5" selects CPUs 1,
 222         3, 4, and 5.  The specified CPUs will be offloaded in addition to
 223         any CPUs specified as offloaded by CONFIG_RCU_NOCB_CPU_ZERO=y or
 224         CONFIG_RCU_NOCB_CPU_ALL=y.  This means that the "rcu_nocbs=" boot
 225         parameter has no effect for kernels built with RCU_NOCB_CPU_ALL=y.
 226
 227 The offloaded CPUs will never queue RCU callbacks, and therefore RCU
 228 never prevents offloaded CPUs from entering either dyntick-idle mode
 229 or adaptive-tick mode.  That said, note that it is up to userspace to
 230 pin the "rcuo" kthreads to specific CPUs if desired.  Otherwise, the
 231 scheduler will decide where to run them, which might or might not be
 232 where you want them to run.
 233
 234
 235 KNOWN ISSUES
 236
 237 o       Dyntick-idle slows transitions to and from idle slightly.
 238         In practice, this has not been a problem except for the most
 239         aggressive real-time workloads, which have the option of disabling
 240         dyntick-idle mode, an option that most of them take.  However,
 241         some workloads will no doubt want to use adaptive ticks to
 242         eliminate scheduling-clock interrupt latencies.  Here are some
 243         options for these workloads:
 244
 245         a.      Use PMQOS from userspace to inform the kernel of your
 246                 latency requirements (preferred).
 247
 248         b.      On x86 systems, use the "idle=mwait" boot parameter.
 249
 250         c.      On x86 systems, use the "intel_idle.max_cstate=" to limit
 251         `       the maximum C-state depth.
 252
 253         d.      On x86 systems, use the "idle=poll" boot parameter.
 254                 However, please note that use of this parameter can cause
 255                 your CPU to overheat, which may cause thermal throttling
 256                 to degrade your latencies -- and that this degradation can
 257                 be even worse than that of dyntick-idle.  Furthermore,
 258                 this parameter effectively disables Turbo Mode on Intel
 259                 CPUs, which can significantly reduce maximum performance.
 260
 261 o       Adaptive-ticks slows user/kernel transitions slightly.
 262         This is not expected to be a problem for computationally intensive
 263         workloads, which have few such transitions.  Careful benchmarking
 264         will be required to determine whether or not other workloads
 265         are significantly affected by this effect.
 266
 267 o       Adaptive-ticks does not do anything unless there is only one
 268         runnable task for a given CPU, even though there are a number
 269         of other situations where the scheduling-clock tick is not
 270         needed.  To give but one example, consider a CPU that has one
 271         runnable high-priority SCHED_FIFO task and an arbitrary number
 272         of low-priority SCHED_OTHER tasks.  In this case, the CPU is
 273         required to run the SCHED_FIFO task until it either blocks or
 274         some other higher-priority task awakens on (or is assigned to)
 275         this CPU, so there is no point in sending a scheduling-clock
 276         interrupt to this CPU.  However, the current implementation
 277         nevertheless sends scheduling-clock interrupts to CPUs having a
 278         single runnable SCHED_FIFO task and multiple runnable SCHED_OTHER
 279         tasks, even though these interrupts are unnecessary.
 280
 281         And even when there are multiple runnable tasks on a given CPU,
 282         there is little point in interrupting that CPU until the current
 283         running task's timeslice expires, which is almost always way
 284         longer than the time of the next scheduling-clock interrupt.
 285
 286         Better handling of these sorts of situations is future work.
 287
 288 o       A reboot is required to reconfigure both adaptive idle and RCU
 289         callback offloading.  Runtime reconfiguration could be provided
 290         if needed, however, due to the complexity of reconfiguring RCU at
 291         runtime, there would need to be an earthshakingly good reason.
 292         Especially given that you have the straightforward option of
 293         simply offloading RCU callbacks from all CPUs and pinning them
 294         where you want them whenever you want them pinned.
 295
 296 o       Additional configuration is required to deal with other sources
 297         of OS jitter, including interrupts and system-utility tasks
 298         and processes.  This configuration normally involves binding
 299         interrupts and tasks to particular CPUs.
 300
 301 o       Some sources of OS jitter can currently be eliminated only by
 302         constraining the workload.  For example, the only way to eliminate
 303         OS jitter due to global TLB shootdowns is to avoid the unmapping
 304         operations (such as kernel module unload operations) that
 305         result in these shootdowns.  For another example, page faults
 306         and TLB misses can be reduced (and in some cases eliminated) by
 307         using huge pages and by constraining the amount of memory used
 308         by the application.  Pre-faulting the working set can also be
 309         helpful, especially when combined with the mlock() and mlockall()
 310         system calls.
 311
 312 o       Unless all CPUs are idle, at least one CPU must keep the
 313         scheduling-clock interrupt going in order to support accurate
 314         timekeeping.
 315
 316 o       If there might potentially be some adaptive-ticks CPUs, there
 317         will be at least one CPU keeping the scheduling-clock interrupt
 318         going, even if all CPUs are otherwise idle.
 319
 320         Better handling of this situation is ongoing work.
 321
 322 o       Some process-handling operations still require the occasional
 323         scheduling-clock tick.  These operations include calculating CPU
 324         load, maintaining sched average, computing CFS entity vruntime,
 325         computing avenrun, and carrying out load balancing.  They are
 326         currently accommodated by scheduling-clock tick every second
 327         or so.  On-going work will eliminate the need even for these
 328         infrequent scheduling-clock ticks.