kvm: Improving undercommit scenarios

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

 kernel/sched/core.c |   25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
  * It's the caller's job to ensure that the target task struct
  * can't go away on us before we can do any checks.
  *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
  */
 bool __sched yield_to(struct task_struct *p, bool preempt)
 {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

 kernel/sched/core.c |   25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
  * It's the caller's job to ensure that the target task struct
  * can't go away on us before we can do any checks.
  *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
  */
 bool __sched yield_to(struct task_struct *p, bool preempt)
 {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

 again:
 	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

Quoted 57 lines · Ingo Molnar

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

 kernel/sched/core.c |   25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
  * It's the caller's job to ensure that the target task struct
  * can't go away on us before we can do any checks.
  *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
  */
 bool __sched yield_to(struct task_struct *p, bool preempt)
 {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

 again:
 	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Quoted 5 lines

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

I ran the test with kernbench and sysbench again on 32 core mx3850 machine with 32 vcpu guests. Results shows definite improvements.

ebizzy and dbench show similar improvement for 1x overcommit (note that stdev for 1x in dbench is lesser improvemet is now seen at only 20%)

[ all the experiments are taken out of 8 run averages ].

The patches benefit large guest undercommit scenarios, so I believe with large guest performance improvemnt is even significant. [ Chegu Vinod results show performance near to no ple cases ]. Unfortunately I do not have a machine to test larger guest (>32).

Ingo, Please let me know if this is okay to you.

base kernel = 3.8.0-rc4

+-----------+-----------+-----------+------------+-----------+
                kernbench  (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   46.6028     1.8672	    42.4494     1.1390	   8.91234
2x   99.9074     9.1859	    90.4050     2.6131	   9.51121
+-----------+-----------+-----------+------------+-----------+
+-----------+-----------+-----------+------------+-----------+
               sysbench (time in sec lower is better) 
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   18.7402     0.3764	    17.7431     0.3589	   5.32065
2x   13.2238     0.1935	    13.0096     0.3152	   1.61981
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                ebizzy  (records/sec higher is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  2421.9000    19.1801	  5883.1000   112.7243	 142.91259
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                dbench (throughput MB/sec  higher is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  11675.9900   857.4154	 14103.5000   215.8425	  20.79061
+-----------+-----------+-----------+------------+-----------+

Quoted 108 lines · Raghavendra K T

Quoted 57 lines · Ingo Molnar

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

  kernel/sched/core.c |   25 +++++++++++++++++++------
  1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
   * It's the caller's job to ensure that the target task struct
   * can't go away on us before we can do any checks.
   *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
   */
  bool __sched yield_to(struct task_struct *p, bool preempt)
  {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

  again:
  	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Quoted 5 lines

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

I ran the test with kernbench and sysbench again on 32 core mx3850 machine with 32 vcpu guests. Results shows definite improvements.

ebizzy and dbench show similar improvement for 1x overcommit (note that stdev for 1x in dbench is lesser improvemet is now seen at only 20%)

[ all the experiments are taken out of 8 run averages ].

The patches benefit large guest undercommit scenarios, so I believe with large guest performance improvemnt is even significant. [ Chegu Vinod results show performance near to no ple cases ]. Unfortunately I do not have a machine to test larger guest (>32).

Ingo, Please let me know if this is okay to you.

base kernel = 3.8.0-rc4

+-----------+-----------+-----------+------------+-----------+
                 kernbench  (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   46.6028     1.8672	    42.4494     1.1390	   8.91234
2x   99.9074     9.1859	    90.4050     2.6131	   9.51121
+-----------+-----------+-----------+------------+-----------+
+-----------+-----------+-----------+------------+-----------+
                sysbench (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   18.7402     0.3764	    17.7431     0.3589	   5.32065
2x   13.2238     0.1935	    13.0096     0.3152	   1.61981
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                 ebizzy  (records/sec higher is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  2421.9000    19.1801	  5883.1000   112.7243	 142.91259
+-----------+-----------+-----------+------------+-----------+

Quoted 7 lines

+-----------+-----------+-----------+------------+-----------+
                 dbench (throughput MB/sec  higher is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  11675.9900   857.4154	 14103.5000   215.8425	  20.79061
+-----------+-----------+-----------+------------+-----------+

The numbers look pretty convincing, thanks. The workloads were CPU bound most of the time, right?

Quoted 119 lines · Ingo Molnar

Quoted 115 lines · Raghavendra K T

Quoted 57 lines · Ingo Molnar

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

 kernel/sched/core.c |   25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
  * It's the caller's job to ensure that the target task struct
  * can't go away on us before we can do any checks.
  *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
  */
 bool __sched yield_to(struct task_struct *p, bool preempt)
 {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

 again:
 	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Quoted 5 lines

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

I ran the test with kernbench and sysbench again on 32 core mx3850 machine with 32 vcpu guests. Results shows definite improvements.

ebizzy and dbench show similar improvement for 1x overcommit (note that stdev for 1x in dbench is lesser improvemet is now seen at only 20%)

[ all the experiments are taken out of 8 run averages ].

The patches benefit large guest undercommit scenarios, so I believe with large guest performance improvemnt is even significant. [ Chegu Vinod results show performance near to no ple cases ]. Unfortunately I do not have a machine to test larger guest (>32).

Ingo, Please let me know if this is okay to you.

base kernel = 3.8.0-rc4

+-----------+-----------+-----------+------------+-----------+
                kernbench  (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   46.6028     1.8672	    42.4494     1.1390	   8.91234
2x   99.9074     9.1859	    90.4050     2.6131	   9.51121
+-----------+-----------+-----------+------------+-----------+
+-----------+-----------+-----------+------------+-----------+
               sysbench (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   18.7402     0.3764	    17.7431     0.3589	   5.32065
2x   13.2238     0.1935	    13.0096     0.3152	   1.61981
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                ebizzy  (records/sec higher is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  2421.9000    19.1801	  5883.1000   112.7243	 142.91259
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                dbench (throughput MB/sec  higher is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  11675.9900   857.4154	 14103.5000   215.8425	  20.79061
+-----------+-----------+-----------+------------+-----------+

Quoted 2 lines

The numbers look pretty convincing, thanks. The workloads were CPU bound most of the time, right?

Yes. CPU bound most of the time. I also used tmpfs to reduce io overhead (for dbbench).

Quoted 123 lines · Raghavendra K T

Quoted 121 lines · Ingo Molnar

Quoted 108 lines · Raghavendra K T

Quoted 57 lines · Ingo Molnar

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

  kernel/sched/core.c |   25 +++++++++++++++++++------
  1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
   * It's the caller's job to ensure that the target task struct
   * can't go away on us before we can do any checks.
   *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
   */
  bool __sched yield_to(struct task_struct *p, bool preempt)
  {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

  again:
  	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Quoted 5 lines

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

I ran the test with kernbench and sysbench again on 32 core mx3850 machine with 32 vcpu guests. Results shows definite improvements.

ebizzy and dbench show similar improvement for 1x overcommit (note that stdev for 1x in dbench is lesser improvemet is now seen at only 20%)

[ all the experiments are taken out of 8 run averages ].

The patches benefit large guest undercommit scenarios, so I believe with large guest performance improvemnt is even significant. [ Chegu Vinod results show performance near to no ple cases ]. Unfortunately I do not have a machine to test larger guest (>32).

Ingo, Please let me know if this is okay to you.

base kernel = 3.8.0-rc4

+-----------+-----------+-----------+------------+-----------+
                 kernbench  (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   46.6028     1.8672	    42.4494     1.1390	   8.91234
2x   99.9074     9.1859	    90.4050     2.6131	   9.51121
+-----------+-----------+-----------+------------+-----------+
+-----------+-----------+-----------+------------+-----------+
                sysbench (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   18.7402     0.3764	    17.7431     0.3589	   5.32065
2x   13.2238     0.1935	    13.0096     0.3152	   1.61981
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                 ebizzy  (records/sec higher is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  2421.9000    19.1801	  5883.1000   112.7243	 142.91259
+-----------+-----------+-----------+------------+-----------+

Quoted 7 lines

+-----------+-----------+-----------+------------+-----------+
                 dbench (throughput MB/sec  higher is better)
+-----------+-----------+-----------+------------+-----------+
     base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  11675.9900   857.4154	 14103.5000   215.8425	  20.79061
+-----------+-----------+-----------+------------+-----------+

The numbers look pretty convincing, thanks. The workloads were CPU bound most of the time, right?

Quoted 2 lines

Yes. CPU bound most of the time. I also used tmpfs to reduce io overhead (for dbbench).

Ok, cool.

Which tree will this be upstreamed through - the KVM tree? I'd suggest the KVM tree because KVM will be the one exposed to the effects of this change.

Quoted 57 lines · Ingo Molnar

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

 kernel/sched/core.c |   25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
  * It's the caller's job to ensure that the target task struct
  * can't go away on us before we can do any checks.
  *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
  */
 bool __sched yield_to(struct task_struct *p, bool preempt)
 {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

 again:
 	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Quoted 5 lines

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

I ran the test with kernbench and sysbench again on 32 core mx3850 machine with 32 vcpu guests. Results shows definite improvements.

ebizzy and dbench show similar improvement for 1x overcommit (note that stdev for 1x in dbench is lesser improvemet is now seen at only 20%)

[ all the experiments are taken out of 8 run averages ].

The patches benefit large guest undercommit scenarios, so I believe with large guest performance improvemnt is even significant. [ Chegu Vinod results show performance near to no ple cases ].

do not have a machine to test larger guest (>32).

Ingo, Please let me know if this is okay to you.

base kernel = 3.8.0-rc4

+-----------+-----------+-----------+------------+-----------+
                kernbench  (time in sec lower is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   46.6028     1.8672	    42.4494     1.1390	   8.91234
2x   99.9074     9.1859	    90.4050     2.6131	   9.51121
+-----------+-----------+-----------+------------+-----------+
+-----------+-----------+-----------+------------+-----------+
               sysbench (time in sec lower is better) 
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x   18.7402     0.3764	    17.7431     0.3589	   5.32065
2x   13.2238     0.1935	    13.0096     0.3152	   1.61981
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                ebizzy  (records/sec higher is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  2421.9000    19.1801	  5883.1000   112.7243	 142.91259
+-----------+-----------+-----------+------------+-----------+

+-----------+-----------+-----------+------------+-----------+
                dbench (throughput MB/sec  higher is better)
+-----------+-----------+-----------+------------+-----------+
    base        stdev        patched    stdev      %improve
+-----------+-----------+-----------+------------+-----------+
1x  11675.9900   857.4154	 14103.5000   215.8425	  20.79061
+-----------+-----------+-----------+------------+-----------+

Quoted 74 lines · Raghavendra K T

Quoted 57 lines · Ingo Molnar

Quoted 41 lines · Raghavendra K T

From: Peter Zijlstra <[email protected]>

In case of undercomitted scenarios, especially in large guests yield_to overhead is significantly high. when run queue length of source and target is one, take an opportunity to bail out and return -ESRCH. This return condition can be further exploited to quickly come out of PLE handler.

(History: Raghavendra initially worked on break out of kvm ple handler upon
seeing source runqueue length = 1, but it had to export rq length). Peter came up with the elegant idea of return -ESRCH in scheduler core.

Signed-off-by Peter Zijlstra <[email protected]>

Raghavendra, Checking the rq length of target vcpu condition added.(thanks Avi)

Reviewed-by Srikar Dronamraju <[email protected]>

Signed-off-by Raghavendra K T <[email protected]>

Acked-by Andrew Jones <[email protected]>

Tested-by Chegu Vinod <[email protected]>

---

  kernel/sched/core.c |   25 +++++++++++++++++++------
  1 file changed, 19 insertions(+), 6 deletions(-)

kernel/sched/core.c +4 -1

index 2d8927f..fc219a5 100644
@@ -4289,7 +4289,10 @@ EXPORT_SYMBOL(yield);
   * It's the caller's job to ensure that the target task struct
   * can't go away on us before we can do any checks.
   *
- * Returns true if we indeed boosted the target task.
+ * Returns:
+ *	true (>0) if we indeed boosted the target task.
+ *	false (0) if we failed to boost the target.
+ *	-ESRCH if there's no task to yield to.
   */
  bool __sched yield_to(struct task_struct *p, bool preempt)
  {

@@ -4303,6 +4306,15 @@ bool __sched yield_to(struct task_struct *p, bool preempt)
 

kernel/sched/core.c bool __sched yield_to(struct task_struct *p, bool preempt) · line 4316

Quoted 10 lines

kernel/sched/core.c +8 -0

  again:
  	p_rq = task_rq(p);
+	/*
+	 * If we're the only runnable task on the rq and target rq also
+	 * has only one task, there's absolutely no point in yielding.
+	 */
+	if (rq->nr_running == 1 && p_rq->nr_running == 1) {
+		yielded = -ESRCH;
+		goto out_irq;
+	}

Looks good to me in principle.

Quoted 5 lines

Would be nice to get more consistent benchmark numbers. Once those are unambiguously showing that this is a win:

Acked-by Ingo Molnar <[email protected]>

I ran the test with kernbench and sysbench again on 32 core mx3850 machine with 32 vcpu guests. Results shows definite improvements.

ebizzy and dbench show similar improvement for 1x overcommit (note that stdev for 1x in dbench is lesser improvemet is now seen at only 20%)

[ all the experiments are taken out of 8 run averages ].

Quoted 3 lines

The patches benefit large guest undercommit scenarios, so I believe with large guest performance improvemnt is even significant. [ Chegu Vinod results show performance near to no ple cases ].

The last results you posted for dbench for the patched 1x case were showing much better throughput than the no-ple 1x case, which is what was strange. Is that still happening? You don't have the no-ple 1x data here this time. The percent errors look a lot better.

In some special scenarios like #vcpu <= #pcpu, PLE handler may
prove very costly, because there is no need to iterate over vcpus and do unsuccessful yield_to burning CPU.

The first patch optimizes all the yield_to by bailing out when there is no need to continue in yield_to (i.e., when there is only one task in source and target rq).

Second patch uses that in PLE handler. Further when a yield_to fails we do not immediately go out of PLE handler instead we try thrice to have better statistical possibility of false return. Otherwise that would affect moderate overcommit cases.

Result on 3.7.0-rc6 kernel shows around 140% improvement for ebizzy 1x and around 51% for dbench 1x with 32 core PLE machine with 32 vcpu guest.

base = 3.7.0-rc6
machine: 32 core mx3850 x5 PLE mc

--+-----------+-----------+-----------+------------+-----------+
ebizzy (rec/sec higher is beter)
--+-----------+-----------+-----------+------------+-----------+
base stdev patched stdev %improve
--+-----------+-----------+-----------+------------+-----------+

1x   2511.3000    21.5409    6051.8000   170.2592   140.98276   
2x   2679.4000   332.4482    2692.3000   251.4005     0.48145
3x   2253.5000   266.4243    2192.1667   178.9753    -2.72169

--+-----------+-----------+-----------+------------+-----------+

--+-----------+-----------+-----------+------------+-----------+
dbench (throughput in MB/sec. higher is better)
--+-----------+-----------+-----------+------------+-----------+
base stdev patched stdev %improve
--+-----------+-----------+-----------+------------+-----------+

1x  6677.4080   638.5048    10098.0060   3449.7026     51.22643
2x  2012.6760    64.7642    2019.0440     62.6702       0.31639
3x  1302.0783    40.8336    1292.7517     27.0515      -0.71629

--+-----------+-----------+-----------+------------+-----------+

Here is the refernce of no ple result.
ebizzy-1x_nople 7592.6000 rec/sec
dbench_1x_nople 7853.6960 MB/sec

The result says we can still improve by 60% for ebizzy, but overall we are getting impressive performance with the patches.

Changes Since V2:
- Dropped global measures usage patch (Peter Zilstra)
- Do not bail out on first failure (Avi Kivity)
- Try thrice for the failure of yield_to to get statistically more correct behaviour.

Changes since V1:
- Discard the idea of exporting nrrunning and optimize in core scheduler (Peter)
- Use yield() instead of schedule in overcommit scenarios (Rik)
- Use loadavg knowledge to detect undercommit/overcommit

Peter Zijlstra (1):
Bail out of yield_to when source and target runqueue has one task

Raghavendra K T (1):
Handle yield_to failure return for potential undercommit case

Please let me know your comments and suggestions.

Link for the discussion of V3 original:
https://lkml.org/lkml/2012/11/26/166

Link for V2:
https://lkml.org/lkml/2012/10/29/287

Link for V1:
https://lkml.org/lkml/2012/9/21/168

 kernel/sched/core.c | 25 +++++++++++++++++++------
 virt/kvm/kvm_main.c | 26 ++++++++++++++++----------
 2 files changed, 35 insertions(+), 16 deletions(-)

Quoted 40 lines · Raghavendra K T

In some special scenarios like #vcpu <= #pcpu, PLE handler may
prove very costly, because there is no need to iterate over vcpus and do unsuccessful yield_to burning CPU.

The first patch optimizes all the yield_to by bailing out when there is no need to continue in yield_to (i.e., when there is only one task in source and target rq).

Second patch uses that in PLE handler. Further when a yield_to fails we do not immediately go out of PLE handler instead we try thrice to have better statistical possibility of false return. Otherwise that would affect moderate overcommit cases.

Result on 3.7.0-rc6 kernel shows around 140% improvement for ebizzy 1x and around 51% for dbench 1x with 32 core PLE machine with 32 vcpu guest.

base = 3.7.0-rc6
machine: 32 core mx3850 x5 PLE mc

--+-----------+-----------+-----------+------------+-----------+
ebizzy (rec/sec higher is beter)
--+-----------+-----------+-----------+------------+-----------+
base stdev patched stdev %improve
--+-----------+-----------+-----------+------------+-----------+

1x   2511.3000    21.5409    6051.8000   170.2592   140.98276
2x   2679.4000   332.4482    2692.3000   251.4005     0.48145
3x   2253.5000   266.4243    2192.1667   178.9753    -2.72169

--+-----------+-----------+-----------+------------+-----------+

--+-----------+-----------+-----------+------------+-----------+
dbench (throughput in MB/sec. higher is better)
--+-----------+-----------+-----------+------------+-----------+
base stdev patched stdev %improve
--+-----------+-----------+-----------+------------+-----------+

1x  6677.4080   638.5048    10098.0060   3449.7026     51.22643
2x  2012.6760    64.7642    2019.0440     62.6702       0.31639
3x  1302.0783    40.8336    1292.7517     27.0515      -0.71629

--+-----------+-----------+-----------+------------+-----------+

Quoted 3 lines

Here is the refernce of no ple result.
ebizzy-1x_nople 7592.6000 rec/sec
dbench_1x_nople 7853.6960 MB/sec

I'm not sure how much we should trust ebizzy results,

so, the dbench results are stranger. The percent error is huge (34%) and somehow we do much better for 1x overcommit with PLE enabled then without (for the patched version). How does that happen? How many guests are running in the 1x test?

throughput results the combined throughput of all of them? I wonder if this jump in throughput is just the guests' perceived throughput, but wrong due to bad virtual time keeping. Can we run a long-lasting benchmark and measure the elapsed time with a clock external from the guests?

In some special scenarios like #vcpu <= #pcpu, PLE handler may
prove very costly, because there is no need to iterate over vcpus and do unsuccessful yield_to burning CPU.

The first patch optimizes all the yield_to by bailing out when there is no need to continue in yield_to (i.e., when there is only one task in source and target rq).

Second patch uses that in PLE handler. Further when a yield_to fails we do not immediately go out of PLE handler instead we try thrice to have better statistical possibility of false return. Otherwise that would affect moderate overcommit cases.

Result on 3.7.0-rc6 kernel shows around 140% improvement for ebizzy 1x and around 51% for dbench 1x with 32 core PLE machine with 32 vcpu guest.

base = 3.7.0-rc6
machine: 32 core mx3850 x5 PLE mc

--+-----------+-----------+-----------+------------+-----------+
ebizzy (rec/sec higher is beter)
--+-----------+-----------+-----------+------------+-----------+
base stdev patched stdev %improve
--+-----------+-----------+-----------+------------+-----------+

1x   2511.3000    21.5409    6051.8000   170.2592   140.98276   
2x   2679.4000   332.4482    2692.3000   251.4005     0.48145
3x   2253.5000   266.4243    2192.1667   178.9753    -2.72169

--+-----------+-----------+-----------+------------+-----------+

--+-----------+-----------+-----------+------------+-----------+
dbench (throughput in MB/sec. higher is better)
--+-----------+-----------+-----------+------------+-----------+
base stdev patched stdev %improve
--+-----------+-----------+-----------+------------+-----------+

1x  6677.4080   638.5048    10098.0060   3449.7026     51.22643
2x  2012.6760    64.7642    2019.0440     62.6702       0.31639
3x  1302.0783    40.8336    1292.7517     27.0515      -0.71629

--+-----------+-----------+-----------+------------+-----------+

Here is the refernce of no ple result.
ebizzy-1x_nople 7592.6000 rec/sec
dbench_1x_nople 7853.6960 MB/sec

The result says we can still improve by 60% for ebizzy, but overall we are getting impressive performance with the patches.

Changes Since V2:
- Dropped global measures usage patch (Peter Zilstra)
- Do not bail out on first failure (Avi Kivity)
- Try thrice for the failure of yield_to to get statistically more correct behaviour.

Changes since V1:
- Discard the idea of exporting nrrunning and optimize in core scheduler (Peter)
- Use yield() instead of schedule in overcommit scenarios (Rik)
- Use loadavg knowledge to detect undercommit/overcommit

Peter Zijlstra (1):
Bail out of yield_to when source and target runqueue has one task

Raghavendra K T (1):
Handle yield_to failure return for potential undercommit case

Please let me know your comments and suggestions.

Link for the discussion of V3 original:
https://lkml.org/lkml/2012/11/26/166

Link for V2:
https://lkml.org/lkml/2012/10/29/287

Link for V1:
https://lkml.org/lkml/2012/9/21/168

 kernel/sched/core.c | 25 +++++++++++++++++++------
 virt/kvm/kvm_main.c | 26 ++++++++++++++++----------
 2 files changed, 35 insertions(+), 16 deletions(-)