247 lines · plain
1=====================2CFS Bandwidth Control3=====================4 5.. note::6 This document only discusses CPU bandwidth control for SCHED_NORMAL.7 The SCHED_RT case is covered in Documentation/scheduler/sched-rt-group.rst8 9CFS bandwidth control is a CONFIG_FAIR_GROUP_SCHED extension which allows the10specification of the maximum CPU bandwidth available to a group or hierarchy.11 12The bandwidth allowed for a group is specified using a quota and period. Within13each given "period" (microseconds), a task group is allocated up to "quota"14microseconds of CPU time. That quota is assigned to per-cpu run queues in15slices as threads in the cgroup become runnable. Once all quota has been16assigned any additional requests for quota will result in those threads being17throttled. Throttled threads will not be able to run again until the next18period when the quota is replenished.19 20A group's unassigned quota is globally tracked, being refreshed back to21cfs_quota units at each period boundary. As threads consume this bandwidth it22is transferred to cpu-local "silos" on a demand basis. The amount transferred23within each of these updates is tunable and described as the "slice".24 25Burst feature26-------------27This feature borrows time now against our future underrun, at the cost of28increased interference against the other system users. All nicely bounded.29 30Traditional (UP-EDF) bandwidth control is something like:31 32 (U = \Sum u_i) <= 133 34This guaranteeds both that every deadline is met and that the system is35stable. After all, if U were > 1, then for every second of walltime,36we'd have to run more than a second of program time, and obviously miss37our deadline, but the next deadline will be further out still, there is38never time to catch up, unbounded fail.39 40The burst feature observes that a workload doesn't always executes the full41quota; this enables one to describe u_i as a statistical distribution.42 43For example, have u_i = {x,e}_i, where x is the p(95) and x+e p(100)44(the traditional WCET). This effectively allows u to be smaller,45increasing the efficiency (we can pack more tasks in the system), but at46the cost of missing deadlines when all the odds line up. However, it47does maintain stability, since every overrun must be paired with an48underrun as long as our x is above the average.49 50That is, suppose we have 2 tasks, both specify a p(95) value, then we51have a p(95)*p(95) = 90.25% chance both tasks are within their quota and52everything is good. At the same time we have a p(5)p(5) = 0.25% chance53both tasks will exceed their quota at the same time (guaranteed deadline54fail). Somewhere in between there's a threshold where one exceeds and55the other doesn't underrun enough to compensate; this depends on the56specific CDFs.57 58At the same time, we can say that the worst case deadline miss, will be59\Sum e_i; that is, there is a bounded tardiness (under the assumption60that x+e is indeed WCET).61 62The interferenece when using burst is valued by the possibilities for63missing the deadline and the average WCET. Test results showed that when64there many cgroups or CPU is under utilized, the interference is65limited. More details are shown in:66https://lore.kernel.org/lkml/5371BD36-55AE-4F71-B9D7-B86DC32E3D2B@linux.alibaba.com/67 68Management69----------70Quota, period and burst are managed within the cpu subsystem via cgroupfs.71 72.. note::73 The cgroupfs files described in this section are only applicable74 to cgroup v1. For cgroup v2, see75 :ref:`Documentation/admin-guide/cgroup-v2.rst <cgroup-v2-cpu>`.76 77- cpu.cfs_quota_us: run-time replenished within a period (in microseconds)78- cpu.cfs_period_us: the length of a period (in microseconds)79- cpu.stat: exports throttling statistics [explained further below]80- cpu.cfs_burst_us: the maximum accumulated run-time (in microseconds)81 82The default values are::83 84 cpu.cfs_period_us=100ms85 cpu.cfs_quota_us=-186 cpu.cfs_burst_us=087 88A value of -1 for cpu.cfs_quota_us indicates that the group does not have any89bandwidth restriction in place, such a group is described as an unconstrained90bandwidth group. This represents the traditional work-conserving behavior for91CFS.92 93Writing any (valid) positive value(s) no smaller than cpu.cfs_burst_us will94enact the specified bandwidth limit. The minimum quota allowed for the quota or95period is 1ms. There is also an upper bound on the period length of 1s.96Additional restrictions exist when bandwidth limits are used in a hierarchical97fashion, these are explained in more detail below.98 99Writing any negative value to cpu.cfs_quota_us will remove the bandwidth limit100and return the group to an unconstrained state once more.101 102A value of 0 for cpu.cfs_burst_us indicates that the group can not accumulate103any unused bandwidth. It makes the traditional bandwidth control behavior for104CFS unchanged. Writing any (valid) positive value(s) no larger than105cpu.cfs_quota_us into cpu.cfs_burst_us will enact the cap on unused bandwidth106accumulation.107 108Any updates to a group's bandwidth specification will result in it becoming109unthrottled if it is in a constrained state.110 111System wide settings112--------------------113For efficiency run-time is transferred between the global pool and CPU local114"silos" in a batch fashion. This greatly reduces global accounting pressure115on large systems. The amount transferred each time such an update is required116is described as the "slice".117 118This is tunable via procfs::119 120 /proc/sys/kernel/sched_cfs_bandwidth_slice_us (default=5ms)121 122Larger slice values will reduce transfer overheads, while smaller values allow123for more fine-grained consumption.124 125Statistics126----------127A group's bandwidth statistics are exported via 5 fields in cpu.stat.128 129cpu.stat:130 131- nr_periods: Number of enforcement intervals that have elapsed.132- nr_throttled: Number of times the group has been throttled/limited.133- throttled_time: The total time duration (in nanoseconds) for which entities134 of the group have been throttled.135- nr_bursts: Number of periods burst occurs.136- burst_time: Cumulative wall-time (in nanoseconds) that any CPUs has used137 above quota in respective periods.138 139This interface is read-only.140 141Hierarchical considerations142---------------------------143The interface enforces that an individual entity's bandwidth is always144attainable, that is: max(c_i) <= C. However, over-subscription in the145aggregate case is explicitly allowed to enable work-conserving semantics146within a hierarchy:147 148 e.g. \Sum (c_i) may exceed C149 150[ Where C is the parent's bandwidth, and c_i its children ]151 152 153There are two ways in which a group may become throttled:154 155 a. it fully consumes its own quota within a period156 b. a parent's quota is fully consumed within its period157 158In case b) above, even though the child may have runtime remaining it will not159be allowed to until the parent's runtime is refreshed.160 161CFS Bandwidth Quota Caveats162---------------------------163Once a slice is assigned to a cpu it does not expire. However all but 1ms of164the slice may be returned to the global pool if all threads on that cpu become165unrunnable. This is configured at compile time by the min_cfs_rq_runtime166variable. This is a performance tweak that helps prevent added contention on167the global lock.168 169The fact that cpu-local slices do not expire results in some interesting corner170cases that should be understood.171 172For cgroup cpu constrained applications that are cpu limited this is a173relatively moot point because they will naturally consume the entirety of their174quota as well as the entirety of each cpu-local slice in each period. As a175result it is expected that nr_periods roughly equal nr_throttled, and that176cpuacct.usage will increase roughly equal to cfs_quota_us in each period.177 178For highly-threaded, non-cpu bound applications this non-expiration nuance179allows applications to briefly burst past their quota limits by the amount of180unused slice on each cpu that the task group is running on (typically at most1811ms per cpu or as defined by min_cfs_rq_runtime). This slight burst only182applies if quota had been assigned to a cpu and then not fully used or returned183in previous periods. This burst amount will not be transferred between cores.184As a result, this mechanism still strictly limits the task group to quota185average usage, albeit over a longer time window than a single period. This186also limits the burst ability to no more than 1ms per cpu. This provides187better more predictable user experience for highly threaded applications with188small quota limits on high core count machines. It also eliminates the189propensity to throttle these applications while simultaneously using less than190quota amounts of cpu. Another way to say this, is that by allowing the unused191portion of a slice to remain valid across periods we have decreased the192possibility of wastefully expiring quota on cpu-local silos that don't need a193full slice's amount of cpu time.194 195The interaction between cpu-bound and non-cpu-bound-interactive applications196should also be considered, especially when single core usage hits 100%. If you197gave each of these applications half of a cpu-core and they both got scheduled198on the same CPU it is theoretically possible that the non-cpu bound application199will use up to 1ms additional quota in some periods, thereby preventing the200cpu-bound application from fully using its quota by that same amount. In these201instances it will be up to the CFS algorithm (see sched-design-CFS.rst) to202decide which application is chosen to run, as they will both be runnable and203have remaining quota. This runtime discrepancy will be made up in the following204periods when the interactive application idles.205 206Examples207--------2081. Limit a group to 1 CPU worth of runtime::209 210 If period is 250ms and quota is also 250ms, the group will get211 1 CPU worth of runtime every 250ms.212 213 # echo 250000 > cpu.cfs_quota_us /* quota = 250ms */214 # echo 250000 > cpu.cfs_period_us /* period = 250ms */215 2162. Limit a group to 2 CPUs worth of runtime on a multi-CPU machine217 218 With 500ms period and 1000ms quota, the group can get 2 CPUs worth of219 runtime every 500ms::220 221 # echo 1000000 > cpu.cfs_quota_us /* quota = 1000ms */222 # echo 500000 > cpu.cfs_period_us /* period = 500ms */223 224 The larger period here allows for increased burst capacity.225 2263. Limit a group to 20% of 1 CPU.227 228 With 50ms period, 10ms quota will be equivalent to 20% of 1 CPU::229 230 # echo 10000 > cpu.cfs_quota_us /* quota = 10ms */231 # echo 50000 > cpu.cfs_period_us /* period = 50ms */232 233 By using a small period here we are ensuring a consistent latency234 response at the expense of burst capacity.235 2364. Limit a group to 40% of 1 CPU, and allow accumulate up to 20% of 1 CPU237 additionally, in case accumulation has been done.238 239 With 50ms period, 20ms quota will be equivalent to 40% of 1 CPU.240 And 10ms burst will be equivalent to 20% of 1 CPU::241 242 # echo 20000 > cpu.cfs_quota_us /* quota = 20ms */243 # echo 50000 > cpu.cfs_period_us /* period = 50ms */244 # echo 10000 > cpu.cfs_burst_us /* burst = 10ms */245 246 Larger buffer setting (no larger than quota) allows greater burst capacity.247