qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Cc: Utkarsh Tripathi <utkarsh.tripathi@nutanix.com>
Subject: [PULL 36/46] cpu-throttle: Remove timer_mod() from cpu_throttle_set()
Date: Mon,  8 Feb 2021 19:23:21 +0100	[thread overview]
Message-ID: <20210208182331.58897-37-pbonzini@redhat.com> (raw)
In-Reply-To: <20210208182331.58897-1-pbonzini@redhat.com>

From: Utkarsh Tripathi <utkarsh.tripathi@nutanix.com>

During migrations, after each iteration, cpu_throttle_set() is called,
which irrespective of input, re-arms the timer according to value of
new_throttle_pct. This causes cpu_throttle_thread() to be delayed in
getting scheduled and consqeuntly lets guest run for more time than what
the throttle value should allow. This leads to spikes in guest throughput
at high cpu-throttle percentage whenever cpu_throttle_set() is called.

A solution would be not to modify the timer immediately in
cpu_throttle_set(), instead, only modify throttle_percentage so that the
throttle would automatically adjust to the required percentage when
cpu_throttle_timer_tick() is invoked.

Manually tested the patch using following configuration:

Guest:
Centos7 (3.10.0-123.el7.x86_64)
Total Memory - 64GB , CPUs - 16
Tool used - stress (1.0.4)
Workload - stress --vm 32 --vm-bytes 1G --vm-keep

Migration Parameters:
Network Bandwidth - 500MBPS
cpu-throttle-initial - 99

Results:
With timer_mod(): fails to converge, continues indefinitely
Without timer_mod(): converges in 249 sec

Signed-off-by: Utkarsh Tripathi <utkarsh.tripathi@nutanix.com>
Message-Id: <1609420384-119407-1-git-send-email-utkarsh.tripathi@nutanix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 softmmu/cpu-throttle.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/softmmu/cpu-throttle.c b/softmmu/cpu-throttle.c
index 2ec4b8e0bc..8c2144ab95 100644
--- a/softmmu/cpu-throttle.c
+++ b/softmmu/cpu-throttle.c
@@ -90,14 +90,21 @@ static void cpu_throttle_timer_tick(void *opaque)
 
 void cpu_throttle_set(int new_throttle_pct)
 {
+    /*
+     * boolean to store whether throttle is already active or not,
+     * before modifying throttle_percentage
+     */
+    bool throttle_active = cpu_throttle_active();
+
     /* Ensure throttle percentage is within valid range */
     new_throttle_pct = MIN(new_throttle_pct, CPU_THROTTLE_PCT_MAX);
     new_throttle_pct = MAX(new_throttle_pct, CPU_THROTTLE_PCT_MIN);
 
     qatomic_set(&throttle_percentage, new_throttle_pct);
 
-    timer_mod(throttle_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL_RT) +
-                                       CPU_THROTTLE_TIMESLICE_NS);
+    if (!throttle_active) {
+        cpu_throttle_timer_tick(NULL);
+    }
 }
 
 void cpu_throttle_stop(void)
-- 
2.29.2




  parent reply	other threads:[~2021-02-08 22:29 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-02-08 18:22 [PULL 00/46] Misc patches for 2021-02-08 Paolo Bonzini
2021-02-08 18:22 ` [PULL 01/46] fuzz: ignore address_space_map is_write flag Paolo Bonzini
2021-02-08 18:22 ` [PULL 02/46] fuzz: refine the ide/ahci fuzzer configs Paolo Bonzini
2021-02-08 18:22 ` [PULL 03/46] docs/fuzz: fix pre-meson path Paolo Bonzini
2021-02-08 18:22 ` [PULL 04/46] fuzz: log the arguments used to initialize QEMU Paolo Bonzini
2021-02-08 18:22 ` [PULL 05/46] fuzz: enable dynamic args for generic-fuzz configs Paolo Bonzini
2021-02-08 18:22 ` [PULL 06/46] docs/fuzz: add some information about OSS-Fuzz Paolo Bonzini
2021-02-08 18:22 ` [PULL 07/46] fuzz: add virtio-9p configurations for fuzzing Paolo Bonzini
2021-02-08 18:22 ` [PULL 08/46] target/i386: do not set LM for 32-bit emulation "-cpu host/max" Paolo Bonzini
2021-02-08 18:22 ` [PULL 09/46] machine: add missing doc for memory-backend option Paolo Bonzini
2021-02-08 18:22 ` [PULL 10/46] meson: accept either shared or static libraries if --disable-static Paolo Bonzini
2021-02-08 18:22 ` [PULL 11/46] meson: honor --enable-rbd if cc.links test fails Paolo Bonzini
2021-02-08 18:22 ` [PULL 12/46] x86/cpu: Populate SVM CPUID feature bits Paolo Bonzini
2021-02-08 18:22 ` [PULL 13/46] fuzz: fix wrong index in clear_bits Paolo Bonzini
2021-02-08 18:22 ` [PULL 14/46] docs: don't install corresponding man page if guest agent is disabled Paolo Bonzini
2021-02-08 18:23 ` [PULL 15/46] virtio-scsi: don't uninitialize queues that we didn't initialize Paolo Bonzini
2021-02-08 18:23 ` [PULL 16/46] event_notifier: handle initialization failure better Paolo Bonzini
2021-02-08 18:23 ` [PULL 17/46] target/i386: Fix decoding of certain BMI instructions Paolo Bonzini
2021-02-08 18:23 ` [PULL 18/46] target/i86: implement PKS Paolo Bonzini
2021-02-08 18:23 ` [PULL 19/46] configure: Improve TCI feature description Paolo Bonzini
2021-02-08 18:23 ` [PULL 20/46] meson: Explicit TCG backend used Paolo Bonzini
2021-02-08 18:23 ` [PULL 21/46] meson: Warn when TCI is selected but TCG backend is available Paolo Bonzini
2021-05-21  8:49   ` Peter Maydell
2021-02-08 18:23 ` [PULL 22/46] tests/meson: Only build softfloat objects if TCG is selected Paolo Bonzini
2021-02-08 18:23 ` [PULL 23/46] pc-bios/meson: Only install EDK2 blob firmwares with system emulation Paolo Bonzini
2021-02-08 18:23 ` [PULL 24/46] meson: Restrict block subsystem processing Paolo Bonzini
2021-02-08 18:23 ` [PULL 25/46] meson: Merge trace_events_subdirs array Paolo Bonzini
2021-02-08 18:23 ` [PULL 26/46] meson: Restrict some trace event directories to user/system emulation Paolo Bonzini
2021-02-08 18:23 ` [PULL 27/46] meson: Restrict emulation code Paolo Bonzini
2021-02-08 18:23 ` [PULL 28/46] qapi/meson: Restrict qdev code to system-mode emulation Paolo Bonzini
2021-02-08 18:23 ` [PULL 29/46] qapi/meson: Remove QMP from user-mode emulation Paolo Bonzini
2021-02-08 18:23 ` [PULL 30/46] qapi/meson: Restrict system-mode specific modules Paolo Bonzini
2021-02-08 18:23 ` [PULL 31/46] qapi/meson: Restrict UI module to system emulation and tools Paolo Bonzini
2021-02-08 18:23 ` [PULL 32/46] accel/kvm/kvm-all: Fix wrong return code handling in dirty log code Paolo Bonzini
2021-02-08 18:23 ` [PULL 33/46] replay: fix replay of the interrupts Paolo Bonzini
2021-02-08 18:23 ` [PULL 34/46] pc-bios/descriptors: fix paths in json files Paolo Bonzini
2021-02-08 18:23 ` [PULL 35/46] replay: rng-builtin support Paolo Bonzini
2021-02-08 18:23 ` Paolo Bonzini [this message]
2021-02-08 18:23 ` [PULL 37/46] hw/pci-host: add pci-intack write method Paolo Bonzini
2021-02-08 18:23 ` [PULL 38/46] pci-host: designware: add pcie-msi read method Paolo Bonzini
2021-02-08 18:23 ` [PULL 39/46] vfio: add quirk device write method Paolo Bonzini
2021-02-08 18:23 ` [PULL 40/46] prep: add ppc-parity " Paolo Bonzini
2021-02-08 18:23 ` [PULL 41/46] nvram: add nrf51_soc flash read method Paolo Bonzini
2021-02-08 18:23 ` [PULL 42/46] spapr_pci: add spapr msi " Paolo Bonzini
2021-02-08 18:23 ` [PULL 43/46] tz-ppc: add dummy read/write methods Paolo Bonzini
2021-02-08 18:23 ` [PULL 44/46] imx7-ccm: add digprog mmio write method Paolo Bonzini
2021-02-08 18:23 ` [PULL 45/46] target/i386: Add support for save/load IA32_PKRS MSR Paolo Bonzini
2021-02-08 18:23 ` [PULL 46/46] target/i386: Expose VMX entry/exit load pkrs control bits Paolo Bonzini
2021-02-09 13:24 ` [PULL 00/46] Misc patches for 2021-02-08 Peter Maydell

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210208182331.58897-37-pbonzini@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=utkarsh.tripathi@nutanix.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).