Messages in this thread | | | Date | Tue, 2 Aug 2022 10:10:10 +0530 | Subject | Re: [PATCH] sched/fair: no sync wakeup from interrupt context | From | K Prateek Nayak <> |
| |
Hello Libo,
Thank you for looking into this.
On 8/1/2022 8:27 PM, Libo Chen wrote: > > > On 7/28/22 21:47, K Prateek Nayak wrote: >> Hello Libo and Peter, >> >> tl;dr >> >> - We observed a major regression with tbench when testing the latest tip >> sched/core at: >> commit 14b3f2d9ee8d "sched/fair: Disallow sync wakeup from interrupt context" >> Reason for the regression are the fewer affine wakeups that leaves the >> client farther away from the data it needs to consume next primed in the >> waker's LLC. >> Such regressions can be expected from tasks that use sockets to communicate >> significant amount of data especially on system with multiple LLCs. >> >> - Other benchmarks have a comparable behavior to the tip at previous commit >> commit : 91caa5ae2424 "sched/core: Fix the bug that task won't enqueue >> into core tree when update cookie" >> >> I'll leave more details below. >> >> On 7/12/2022 4:17 AM, Libo Chen wrote: >>> [..snip..] >> >> The two tests kernels used are: >> >> - tip at commit: 14b3f2d9ee8d "sched/fair: Disallow sync wakeup from interrupt context" >> - tip at commit: 91caa5ae2424 "sched/core: Fix the bug that task won't enqueue into core tree when update cookie" >> >> Below are the tbench result on a dual socket Zen3 machine >> running in NPS1 mode. Following is the NUMA configuration >> NPS1 mode: >> >> - NPS1: Each socket is a NUMA node. >> Total 2 NUMA nodes in the dual socket machine. >> >> Node 0: 0-63, 128-191 >> Node 1: 64-127, 192-255 >> >> Clients: tip (91caa5ae2424) tip (14b3f2d9ee8d) >> 1 569.24 (0.00 pct) 283.63 (-50.17 pct) * >> 2 1104.76 (0.00 pct) 590.45 (-46.55 pct) * >> 4 2058.78 (0.00 pct) 1080.63 (-47.51 pct) * >> 8 3590.20 (0.00 pct) 2098.05 (-41.56 pct) * >> 16 6119.21 (0.00 pct) 4348.40 (-28.93 pct) * >> 32 11383.91 (0.00 pct) 8417.55 (-26.05 pct) * >> 64 21910.01 (0.00 pct) 19405.11 (-11.43 pct) * >> 128 33105.27 (0.00 pct) 29791.80 (-10.00 pct) * >> 256 45550.99 (0.00 pct) 45847.10 (0.65 pct) >> 512 57753.81 (0.00 pct) 49481.17 (-14.32 pct) * >> 1024 55684.33 (0.00 pct) 48748.38 (-12.45 pct) * >> >> This regression is consistently reproducible. > I ran tbench with 1 client on my 8 nodes zen2 machine because 1~4 clients count generally shouldn't be affected by this patch. I do see throughput regresses with the patch but > the latency improves pretty much equally. Furthermore, I also don't see tbench tasks being separated in different LLC domains from my ftrace, they are almost always in the same CCXes. > What I do see is there are a lot less interrupts and context switches, and average CPU frequency is slower too with the patch. This is bizarre that Intel doesn't seem to be impacted. > Trying to understand why right now.
Thank you for analyzing this. I see a drop in max latency with the patch but the average latency have increased on the patched kernel. Following are the logs from one of the runs for 1 client case:
- tip (91caa5ae2424)
Operation Count AvgLat MaxLat -------------------------------------------------- Deltree 28 0.000 0.001 Flush 76361 0.008 0.018 Close 800338 0.008 0.080 LockX 3546 0.008 0.015 Mkdir 14 0.008 0.009 Rename 46131 0.008 0.050 ReadX 1707761 0.009 0.139 WriteX 543274 0.012 0.092 Unlink 220019 0.008 0.083 UnlockX 3546 0.008 0.016 FIND_FIRST 381795 0.008 0.079 SET_FILE_INFORMATION 88740 0.008 0.080 QUERY_FILE_INFORMATION 173062 0.008 0.061 QUERY_PATH_INFORMATION 987524 0.008 0.070 QUERY_FS_INFORMATION 181068 0.008 0.049 NTCreateX 1089543 0.008 0.083
Throughput 570.36 MB/sec 1 clients 1 procs max_latency=0.140 ms
- tip (14b3f2d9ee8d)
Operation Count AvgLat MaxLat -------------------------------------------------- Deltree 14 0.000 0.001 Flush 38993 0.017 0.059 Close 408696 0.017 0.085 LockX 1810 0.017 0.023 Mkdir 7 0.016 0.017 Rename 23555 0.017 0.052 ReadX 871996 0.018 0.097 WriteX 277343 0.025 0.105 Unlink 112357 0.017 0.055 UnlockX 1810 0.017 0.023 FIND_FIRST 194961 0.017 0.089 SET_FILE_INFORMATION 45312 0.017 0.032 QUERY_FILE_INFORMATION 88356 0.017 0.078 QUERY_PATH_INFORMATION 504289 0.017 0.119 QUERY_FS_INFORMATION 92460 0.017 0.085 NTCreateX 556374 0.017 0.097
Throughput 291.163 MB/sec 1 clients 1 procs max_latency=0.119 ms
I had only analyzed the schedstat data which showed a clear shift in the number of affine wakeups. I haven't measured the average CPU frequency during the runs. The numbers reported are with the performance governor. I'll try to get more data on the CPU frequency front.
>> Below are the statistics gathered from schedstat data: >> >> Kernel : tip + remove 14b3f2d9ee8d tip >> HEAD commit : 91caa5ae2424 14b3f2d9ee8d >> sched_yield cnt : 11 17 >> Legacy counter can be ignored : 0 0 >> schedule called : 12621212 15104022 >> schedule left the processor idle : 6306653 ( 49.96% of times ) 7547375 ( 49.96% of times ) >> try_to_wake_up was called : 6310778 7552147 >> try_to_wake_up was called to wake up the local cpu : 12305 ( 0.19% of times ) 12816 ( 0.16% of times ) >> total time by tasks on this processor (in jiffies) : 78497712520 72461915902 >> total time waiting tasks on this processor (in jiffies) : 56398803 ( 0.07% of times ) 34631178 ( 0.04% of times ) >> total timeslices run on this cpu : 6314548 7556630 >> >> Wakeups on same SMT : 39 ( 0.00062 ) 263 ( 0.00348 ) >> Wakeups on same MC : 6297015 ( 99.78% of time ) <--- 1079 ( 0.01429 ) >> Wakeups on same DIE : 559 ( 0.00886 ) 7537909 ( 99.81147 ) <--- With the patch, the task will prefer > I don't have a zen3 right now. What is the span of your MC domain as well as DIE?
on Zen3, a group in MC domain consists of the 16 CPUs on the same CCD. On a dual socket Zen3 system (2 x 64C/128T) running in NPS 1 mode, the DIE domain will consists of all the CPUs on the same socket. There are two DIE groups in the dual socket test system. Following are the span of each:
- DIE0: 0-63,128-191
DIE 0 MC 0: 0-7,128-135 DIE 0 MC 1: 8-15,136-143 DIE 0 MC 2: 16-23,144-151 DIE 0 MC 3: 24-31,152-159 DIE 0 MC 4: 32-39,160-167 DIE 0 MC 5: 40-47,168-175 DIE 0 MC 6: 48-55,176-183 DIE 0 MC 7: 56-63,184-191
- DIE1: 64-127,192-255
DIE 1 MC 0: 64-71,192-199 DIE 1 MC 1: 72-79,200-207 DIE 1 MC 2: 80-87,208-215 DIE 1 MC 3: 88-95,216-223 DIE 1 MC 4: 96-103,224-231 DIE 1 MC 5: 104-111,232-239 DIE 1 MC 6: 112-119,240-247 DIE 1 MC 7: 120-127,248-255 > > Thanks for the testing. > > Libo >> Wakeups on same NUMA : 860 ( 0.01363 ) 80 ( 0.00106 ) to wake on the same LLC where it previously >> Affine wakeups on same SMT : 25 ( 0.00040 ) 255 ( 0.00338 ) ran as compared to the LLC of waker. >> Affine wakeups on same MC : 6282684 ( 99.55% of time ) <--- 961 ( 0.01272 ) This results in performance degradation as >> Affine wakeups on same DIE : 523 ( 0.00829 ) 7537220 ( 99.80235 ) <--- the task is farther away from data it will >> Affine wakeups on same NUMA : 839 ( 0.01329 ) 46 ( 0.00061 ) consume next. >> >> All the statistics are comparable except for the reduced number of affine >> wakeup on the waker's LLC that resulting in task being placed on the previous >> LLC farther away from the data that resides in the waker's LLC that the wakee >> will consume next. >> >> All wakeups in the tbench, happens in_serving_softirq() making in_taks() false >> for all the cases where sync would have been true otherwise. >> >> We wanted to highlight there are workloads which would still benefit from >> affine wakeups even when it happens in an interrupt context. It would be >> great if we could spot such cases and bias wakeups towards waker's LLC. >> >> Other benchmarks results are comparable to the tip in most cases. >> All benchmarks were run on machine configured in NPS1 mode. >> Following are the results: >> >> ~~~~~~~~~ >> hackbench >> ~~~~~~~~~ >> >> Test: tip (91caa5ae2424) tip (14b3f2d9ee8d) >> 1-groups: 4.22 (0.00 pct) 4.48 (-6.16 pct) * >> 1-groups: 4.22 (0.00 pct) 4.30 (-1.89 pct) [Verification run] >> 2-groups: 5.01 (0.00 pct) 4.87 (2.79 pct) >> 4-groups: 5.49 (0.00 pct) 5.34 (2.73 pct) >> 8-groups: 5.64 (0.00 pct) 5.50 (2.48 pct) >> 16-groups: 7.54 (0.00 pct) 7.34 (2.65 pct) >> >> ~~~~~~~~ >> schbench >> ~~~~~~~~ >> >> #workers: tip (91caa5ae2424) tip (14b3f2d9ee8d) >> 1: 22.00 (0.00 pct) 22.00 (0.00 pct) >> 2: 22.00 (0.00 pct) 27.00 (-22.72 pct) * Known to have run to run >> 4: 33.00 (0.00 pct) 38.00 (-15.15 pct) * variations. >> 8: 48.00 (0.00 pct) 51.00 (-6.25 pct) * >> 16: 70.00 (0.00 pct) 70.00 (0.00 pct) >> 32: 118.00 (0.00 pct) 120.00 (-1.69 pct) >> 64: 217.00 (0.00 pct) 223.00 (-2.76 pct) >> 128: 485.00 (0.00 pct) 488.00 (-0.61 pct) >> 256: 1066.00 (0.00 pct) 1054.00 (1.12 pct) >> 512: 47296.00 (0.00 pct) 47168.00 (0.27 pct) >> >> Note: schbench results at lower worker count have a large >> run-to-run variance and depends on certain characteristics >> of new-idle balance. >> >> ~~~~~~ >> stream >> ~~~~~~ >> >> - 10 runs >> >> Test: tip (91caa5ae2424) tip (14b3f2d9ee8d) >> Copy: 336140.45 (0.00 pct) 334362.29 (-0.52 pct) >> Scale: 214679.13 (0.00 pct) 218016.44 (1.55 pct) >> Add: 251691.67 (0.00 pct) 249734.04 (-0.77 pct) >> Triad: 262174.93 (0.00 pct) 260063.57 (-0.80 pct) >> >> - 100 runs >> >> Test: tip (91caa5ae2424) tip (14b3f2d9ee8d) >> Copy: 336576.38 (0.00 pct) 334646.27 (-0.57 pct) >> Scale: 219124.86 (0.00 pct) 223480.29 (1.98 pct) >> Add: 251796.93 (0.00 pct) 250845.95 (-0.37 pct) >> Triad: 262286.47 (0.00 pct) 258020.57 (-1.62 pct) >> >> ~~~~~~~~~~~~ >> ycsb-mongodb >> ~~~~~~~~~~~~ >> >> tip (91caa5ae2424): 290479.00 (var: 1.53) >> tip (14b3f2d9ee8d): 287361.67 (var: 0.80) (-1.07 pct) >> >>> [..snip..] >
Thank you again for looking into this issue and for sharing the observations on the Zen2 system. -- Thanks and Regards, Prateek
| |