Re: [RFC PATCH 2/3] sched: power aware load balance,
On 11/12/2012 02:49 AM, Preeti Murthy wrote: > Hi Alex > I apologise for the delay in replying . That's all right. I often also busy on other Intel tasks and have no time to look at LKML. :) > > On Wed, Nov 7, 2012 at 6:57 PM, Alex Shi wrote: >> On 11/07/2012 12:37 PM, Preeti Murthy wrote: >>> Hi Alex, >>> >>> What I am concerned about in this patchset as Peter also >>> mentioned in the previous discussion of your approach >>> (https://lkml.org/lkml/2012/8/13/139) >>> is that: >>> >>> 1.Using nr_running of two different sched groups to decide which one >>> can be group_leader or group_min might not be be the right approach, >>> as this might mislead us to think that a group running one task is less >>> loaded than the group running three tasks although the former task is >>> a cpu hogger. >>> >>> 2.Comparing the number of cpus with the number of tasks running in a sched >>> group to decide if the group is underloaded or overloaded again faces >>> the same issue.The tasks might be short running,not utilizing cpu much. >> >> Yes, maybe nr task is not the best indicator. But as first step, it can >> approve the proposal is a correct path and worth to try more. >> Considering the old powersaving implement is also judge on nr tasks, and >> my testing result of this. It may be still a option. > Hmm.. will think about this and get back. >>> >>> I also feel before we introduce another side to the scheduler called >>> 'power aware',why not try and see if the current scheduler itself can >>> perform better? We have an opportunity in terms of PJT's patches which >>> can help scheduler make more realistic decisions in load balance.Also >>> since PJT's metric is a statistical one,I believe we could vary it to >>> allow scheduler to do more rigorous or less rigorous power savings. >> >> will study the PJT's approach. >> Actually, current patch set is also a kind of load balance modification, >> right? :) > It is true that this is a different approach,in fact we will require > this approach > to do power savings because PJT's patches introduce a new 'metric' and not a > new > 'approach' in my opinion, to do smarter load balancing,not power aware > load balancing per say.So your patch is surely a step towards power > aware lb.I am just worried about the metric used in it. >>> >>> It is true however that this approach will not try and evacuate nearly idle >>> cpus over to nearly full cpus.That is definitely one of the benefits of your >>> patch,in terms of power savings,but I believe your patch is not making use >>> of the right metric to decide that. >> >> If one sched group just has one task, and another group just has one >> LCPU idle, my patch definitely will pull the task to the nearly full >> sched group. So I didn't understand what you mean 'will not try and >> evacuate nearly idle cpus over to nearly full cpus' > No, by 'this approach' I meant the current load balancer integrated with > the PJT's metric.Your approach does 'evacuate' the nearly idle cpus > over to the nearly full cpus.. Oh, a misunderstand on 'this approach'. :) Anyway, we are all clear about this now. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
Hi Alex I apologise for the delay in replying . On Wed, Nov 7, 2012 at 6:57 PM, Alex Shi wrote: > On 11/07/2012 12:37 PM, Preeti Murthy wrote: >> Hi Alex, >> >> What I am concerned about in this patchset as Peter also >> mentioned in the previous discussion of your approach >> (https://lkml.org/lkml/2012/8/13/139) >> is that: >> >> 1.Using nr_running of two different sched groups to decide which one >> can be group_leader or group_min might not be be the right approach, >> as this might mislead us to think that a group running one task is less >> loaded than the group running three tasks although the former task is >> a cpu hogger. >> >> 2.Comparing the number of cpus with the number of tasks running in a sched >> group to decide if the group is underloaded or overloaded again faces >> the same issue.The tasks might be short running,not utilizing cpu much. > > Yes, maybe nr task is not the best indicator. But as first step, it can > approve the proposal is a correct path and worth to try more. > Considering the old powersaving implement is also judge on nr tasks, and > my testing result of this. It may be still a option. Hmm.. will think about this and get back. >> >> I also feel before we introduce another side to the scheduler called >> 'power aware',why not try and see if the current scheduler itself can >> perform better? We have an opportunity in terms of PJT's patches which >> can help scheduler make more realistic decisions in load balance.Also >> since PJT's metric is a statistical one,I believe we could vary it to >> allow scheduler to do more rigorous or less rigorous power savings. > > will study the PJT's approach. > Actually, current patch set is also a kind of load balance modification, > right? :) It is true that this is a different approach,in fact we will require this approach to do power savings because PJT's patches introduce a new 'metric' and not a new 'approach' in my opinion, to do smarter load balancing,not power aware load balancing per say.So your patch is surely a step towards power aware lb.I am just worried about the metric used in it. >> >> It is true however that this approach will not try and evacuate nearly idle >> cpus over to nearly full cpus.That is definitely one of the benefits of your >> patch,in terms of power savings,but I believe your patch is not making use >> of the right metric to decide that. > > If one sched group just has one task, and another group just has one > LCPU idle, my patch definitely will pull the task to the nearly full > sched group. So I didn't understand what you mean 'will not try and > evacuate nearly idle cpus over to nearly full cpus' No, by 'this approach' I meant the current load balancer integrated with the PJT's metric.Your approach does 'evacuate' the nearly idle cpus over to the nearly full cpus.. Regards Preeti U Murthy -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
Hi Alex I apologise for the delay in replying . On Wed, Nov 7, 2012 at 6:57 PM, Alex Shi alex@intel.com wrote: On 11/07/2012 12:37 PM, Preeti Murthy wrote: Hi Alex, What I am concerned about in this patchset as Peter also mentioned in the previous discussion of your approach (https://lkml.org/lkml/2012/8/13/139) is that: 1.Using nr_running of two different sched groups to decide which one can be group_leader or group_min might not be be the right approach, as this might mislead us to think that a group running one task is less loaded than the group running three tasks although the former task is a cpu hogger. 2.Comparing the number of cpus with the number of tasks running in a sched group to decide if the group is underloaded or overloaded again faces the same issue.The tasks might be short running,not utilizing cpu much. Yes, maybe nr task is not the best indicator. But as first step, it can approve the proposal is a correct path and worth to try more. Considering the old powersaving implement is also judge on nr tasks, and my testing result of this. It may be still a option. Hmm.. will think about this and get back. I also feel before we introduce another side to the scheduler called 'power aware',why not try and see if the current scheduler itself can perform better? We have an opportunity in terms of PJT's patches which can help scheduler make more realistic decisions in load balance.Also since PJT's metric is a statistical one,I believe we could vary it to allow scheduler to do more rigorous or less rigorous power savings. will study the PJT's approach. Actually, current patch set is also a kind of load balance modification, right? :) It is true that this is a different approach,in fact we will require this approach to do power savings because PJT's patches introduce a new 'metric' and not a new 'approach' in my opinion, to do smarter load balancing,not power aware load balancing per say.So your patch is surely a step towards power aware lb.I am just worried about the metric used in it. It is true however that this approach will not try and evacuate nearly idle cpus over to nearly full cpus.That is definitely one of the benefits of your patch,in terms of power savings,but I believe your patch is not making use of the right metric to decide that. If one sched group just has one task, and another group just has one LCPU idle, my patch definitely will pull the task to the nearly full sched group. So I didn't understand what you mean 'will not try and evacuate nearly idle cpus over to nearly full cpus' No, by 'this approach' I meant the current load balancer integrated with the PJT's metric.Your approach does 'evacuate' the nearly idle cpus over to the nearly full cpus.. Regards Preeti U Murthy -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
On 11/12/2012 02:49 AM, Preeti Murthy wrote: Hi Alex I apologise for the delay in replying . That's all right. I often also busy on other Intel tasks and have no time to look at LKML. :) On Wed, Nov 7, 2012 at 6:57 PM, Alex Shi alex@intel.com wrote: On 11/07/2012 12:37 PM, Preeti Murthy wrote: Hi Alex, What I am concerned about in this patchset as Peter also mentioned in the previous discussion of your approach (https://lkml.org/lkml/2012/8/13/139) is that: 1.Using nr_running of two different sched groups to decide which one can be group_leader or group_min might not be be the right approach, as this might mislead us to think that a group running one task is less loaded than the group running three tasks although the former task is a cpu hogger. 2.Comparing the number of cpus with the number of tasks running in a sched group to decide if the group is underloaded or overloaded again faces the same issue.The tasks might be short running,not utilizing cpu much. Yes, maybe nr task is not the best indicator. But as first step, it can approve the proposal is a correct path and worth to try more. Considering the old powersaving implement is also judge on nr tasks, and my testing result of this. It may be still a option. Hmm.. will think about this and get back. I also feel before we introduce another side to the scheduler called 'power aware',why not try and see if the current scheduler itself can perform better? We have an opportunity in terms of PJT's patches which can help scheduler make more realistic decisions in load balance.Also since PJT's metric is a statistical one,I believe we could vary it to allow scheduler to do more rigorous or less rigorous power savings. will study the PJT's approach. Actually, current patch set is also a kind of load balance modification, right? :) It is true that this is a different approach,in fact we will require this approach to do power savings because PJT's patches introduce a new 'metric' and not a new 'approach' in my opinion, to do smarter load balancing,not power aware load balancing per say.So your patch is surely a step towards power aware lb.I am just worried about the metric used in it. It is true however that this approach will not try and evacuate nearly idle cpus over to nearly full cpus.That is definitely one of the benefits of your patch,in terms of power savings,but I believe your patch is not making use of the right metric to decide that. If one sched group just has one task, and another group just has one LCPU idle, my patch definitely will pull the task to the nearly full sched group. So I didn't understand what you mean 'will not try and evacuate nearly idle cpus over to nearly full cpus' No, by 'this approach' I meant the current load balancer integrated with the PJT's metric.Your approach does 'evacuate' the nearly idle cpus over to the nearly full cpus.. Oh, a misunderstand on 'this approach'. :) Anyway, we are all clear about this now. -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
On 11/07/2012 12:37 PM, Preeti Murthy wrote: > Hi Alex, > > What I am concerned about in this patchset as Peter also > mentioned in the previous discussion of your approach > (https://lkml.org/lkml/2012/8/13/139) > is that: > > 1.Using nr_running of two different sched groups to decide which one > can be group_leader or group_min might not be be the right approach, > as this might mislead us to think that a group running one task is less > loaded than the group running three tasks although the former task is > a cpu hogger. > > 2.Comparing the number of cpus with the number of tasks running in a sched > group to decide if the group is underloaded or overloaded again faces > the same issue.The tasks might be short running,not utilizing cpu much. Yes, maybe nr task is not the best indicator. But as first step, it can approve the proposal is a correct path and worth to try more. Considering the old powersaving implement is also judge on nr tasks, and my testing result of this. It may be still a option. > > I also feel before we introduce another side to the scheduler called > 'power aware',why not try and see if the current scheduler itself can > perform better? We have an opportunity in terms of PJT's patches which > can help scheduler make more realistic decisions in load balance.Also > since PJT's metric is a statistical one,I believe we could vary it to > allow scheduler to do more rigorous or less rigorous power savings. will study the PJT's approach. Actually, current patch set is also a kind of load balance modification, right? :) > > It is true however that this approach will not try and evacuate nearly idle > cpus over to nearly full cpus.That is definitely one of the benefits of your > patch,in terms of power savings,but I believe your patch is not making use > of the right metric to decide that. If one sched group just has one task, and another group just has one LCPU idle, my patch definitely will pull the task to the nearly full sched group. So I didn't understand what you mean 'will not try and evacuate nearly idle cpus over to nearly full cpus'. > > IMHO,the appraoch towards power aware scheduler should take the following > steps: > > 1.Make use of PJT's per-entity-load tracking metric to allow scheduler to make > more intelligent decisions in load balancing.Test the performance and power > save > numbers. > > 2.If the above shows some characteristic change in behaviour over the earlier > scheduler,it should be either towards power save or towards performance.If > found > positive towards one of them, try varying the calculation of > per-entity-load to see > if it can lean towards the other behaviour.If it can,then there you > go,you have a > knob to change between policies right there! > > 3.If you don't get enough power savings with the above approach then > add your patchset > to evacuate nearly idle towards nearly busy groups,but by using PJT's metric > to > make the decision. > > What do you think? Will consider this. thanks! > > Regards > Preeti U Murthy > On Tue, Nov 6, 2012 at 6:39 PM, Alex Shi wrote: >> This patch enabled the power aware consideration in load balance. >> >> As mentioned in the power aware scheduler proposal, Power aware >> scheduling has 2 assumptions: >> 1, race to idle is helpful for power saving >> 2, shrink tasks on less sched_groups will reduce power consumption >> >> The first assumption make performance policy take over scheduling when >> system busy. >> The second assumption make power aware scheduling try to move >> disperse tasks into fewer groups until that groups are full of tasks. >> >> This patch reuse lots of Suresh's power saving load balance code. >> Now the general enabling logical is: >> 1, Collect power aware scheduler statistics with performance load >> balance statistics collection. >> 2, if domain is eligible for power load balance do it and forget >> performance load balance, else do performance load balance. >> >> Has tried on my 2 sockets * 4 cores * HT NHM EP machine. >> and 2 sockets * 8 cores * HT SNB EP machine. >> In the following checking, when I is 2/4/8/16, all tasks are >> shrank to run on single core or single socket. >> >> $for ((i=0; i < I; i++)) ; do while true; do : ; done & done >> >> Checking the power consuming with a powermeter on the NHM EP. >> powersaving performance >> I = 2 148w160w >> I = 4 175w181w >> I = 8 207w224w >> I = 16 324w324w >> >> On a SNB laptop(4 cores *HT) >> powersaving performance >> I = 2 28w 35w >> I = 4 38w 52w >> I = 6 44w 54w >> I = 8 56w 56w >> >> On the SNB EP machine, when I = 16, power saved more than 100 Watts. >> >> Also tested the specjbb2005 with jrockit, kbuild, their peak performance >> has no clear change with powersaving policy on all machines. Just >> specjbb2005 with openjdk has about 2% drop on NHM EP machine with >> powersaving
Re: [RFC PATCH 2/3] sched: power aware load balance,
On 11/07/2012 03:51 AM, Andrew Morton wrote: > On Tue, 6 Nov 2012 21:09:58 +0800 > Alex Shi wrote: > >> $for ((i=0; i < I; i++)) ; do while true; do : ; done & done >> >> Checking the power consuming with a powermeter on the NHM EP. >> powersaving performance >> I = 2 148w160w >> I = 4 175w181w >> I = 8 207w224w >> I = 16 324w324w >> >> On a SNB laptop(4 cores *HT) >> powersaving performance >> I = 2 28w 35w >> I = 4 38w 52w >> I = 6 44w 54w >> I = 8 56w 56w >> >> On the SNB EP machine, when I = 16, power saved more than 100 Watts. > > Confused. According to the above table, at I=16 the EP machine saved 0 > watts. Typo in the data? Not typo, since the LCPU number in the EP machine is 16, so if I = 16, the powersaving policy doesn't work actually. That is the patch designed for race to idle assumption. The result looks same as the third patch(for fork/exec/wu) applied. Result put here because it is from this patch. > > > Also, that's a pretty narrow test - it's doing fork and exec at very > high frequency and things such as task placement decisions at process > startup might be affecting the results. Also, the load will be quite > kernel-intensive, as opposed to the more typical userspace-intensive > loads. Sorry, why you think it keep do fork/exec? It just generate several 'bash' task to burn CPU, without fork/exec. with I = 8, on my 32 LCPU SNB EP machine: No do_fork calling in 5 seconds. $ sudo perf stat -e probe:* -a sleep 5 Performance counter stats for 'sleep 5': 3 probe:do_execve [100.00%] 0 probe:do_fork [100.00%] And it is not kernel-intensive, it nearly running all in user level. 'Top' output: 25:0%us VS 0.0%sy Tasks: 319 total, 9 running, 310 sleeping, 0 stopped, 0 zombie Cpu(s): 25.0%us, 0.0%sy, 0.0%ni, 74.5%id, 0.4%wa, 0.1%hi, 0.0%si, 0.0%st ... > So, please run a broader set of tests so we can see the effects? > Really, I have no more ideas for the suitable benchmarks. Just tried the kbuild -j 16 on the 32 LCPU SNB EP, power just saved 10%, but compile time increase about ~15%. Seems if the task number is variation around the powersaving criteria number, that just cause trouble. -- Thanks Alex -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
On 11/07/2012 03:51 AM, Andrew Morton wrote: On Tue, 6 Nov 2012 21:09:58 +0800 Alex Shi alex@intel.com wrote: $for ((i=0; i I; i++)) ; do while true; do : ; done done Checking the power consuming with a powermeter on the NHM EP. powersaving performance I = 2 148w160w I = 4 175w181w I = 8 207w224w I = 16 324w324w On a SNB laptop(4 cores *HT) powersaving performance I = 2 28w 35w I = 4 38w 52w I = 6 44w 54w I = 8 56w 56w On the SNB EP machine, when I = 16, power saved more than 100 Watts. Confused. According to the above table, at I=16 the EP machine saved 0 watts. Typo in the data? Not typo, since the LCPU number in the EP machine is 16, so if I = 16, the powersaving policy doesn't work actually. That is the patch designed for race to idle assumption. The result looks same as the third patch(for fork/exec/wu) applied. Result put here because it is from this patch. Also, that's a pretty narrow test - it's doing fork and exec at very high frequency and things such as task placement decisions at process startup might be affecting the results. Also, the load will be quite kernel-intensive, as opposed to the more typical userspace-intensive loads. Sorry, why you think it keep do fork/exec? It just generate several 'bash' task to burn CPU, without fork/exec. with I = 8, on my 32 LCPU SNB EP machine: No do_fork calling in 5 seconds. $ sudo perf stat -e probe:* -a sleep 5 Performance counter stats for 'sleep 5': 3 probe:do_execve [100.00%] 0 probe:do_fork [100.00%] And it is not kernel-intensive, it nearly running all in user level. 'Top' output: 25:0%us VS 0.0%sy Tasks: 319 total, 9 running, 310 sleeping, 0 stopped, 0 zombie Cpu(s): 25.0%us, 0.0%sy, 0.0%ni, 74.5%id, 0.4%wa, 0.1%hi, 0.0%si, 0.0%st ... So, please run a broader set of tests so we can see the effects? Really, I have no more ideas for the suitable benchmarks. Just tried the kbuild -j 16 on the 32 LCPU SNB EP, power just saved 10%, but compile time increase about ~15%. Seems if the task number is variation around the powersaving criteria number, that just cause trouble. -- Thanks Alex -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
On 11/07/2012 12:37 PM, Preeti Murthy wrote: Hi Alex, What I am concerned about in this patchset as Peter also mentioned in the previous discussion of your approach (https://lkml.org/lkml/2012/8/13/139) is that: 1.Using nr_running of two different sched groups to decide which one can be group_leader or group_min might not be be the right approach, as this might mislead us to think that a group running one task is less loaded than the group running three tasks although the former task is a cpu hogger. 2.Comparing the number of cpus with the number of tasks running in a sched group to decide if the group is underloaded or overloaded again faces the same issue.The tasks might be short running,not utilizing cpu much. Yes, maybe nr task is not the best indicator. But as first step, it can approve the proposal is a correct path and worth to try more. Considering the old powersaving implement is also judge on nr tasks, and my testing result of this. It may be still a option. I also feel before we introduce another side to the scheduler called 'power aware',why not try and see if the current scheduler itself can perform better? We have an opportunity in terms of PJT's patches which can help scheduler make more realistic decisions in load balance.Also since PJT's metric is a statistical one,I believe we could vary it to allow scheduler to do more rigorous or less rigorous power savings. will study the PJT's approach. Actually, current patch set is also a kind of load balance modification, right? :) It is true however that this approach will not try and evacuate nearly idle cpus over to nearly full cpus.That is definitely one of the benefits of your patch,in terms of power savings,but I believe your patch is not making use of the right metric to decide that. If one sched group just has one task, and another group just has one LCPU idle, my patch definitely will pull the task to the nearly full sched group. So I didn't understand what you mean 'will not try and evacuate nearly idle cpus over to nearly full cpus'. IMHO,the appraoch towards power aware scheduler should take the following steps: 1.Make use of PJT's per-entity-load tracking metric to allow scheduler to make more intelligent decisions in load balancing.Test the performance and power save numbers. 2.If the above shows some characteristic change in behaviour over the earlier scheduler,it should be either towards power save or towards performance.If found positive towards one of them, try varying the calculation of per-entity-load to see if it can lean towards the other behaviour.If it can,then there you go,you have a knob to change between policies right there! 3.If you don't get enough power savings with the above approach then add your patchset to evacuate nearly idle towards nearly busy groups,but by using PJT's metric to make the decision. What do you think? Will consider this. thanks! Regards Preeti U Murthy On Tue, Nov 6, 2012 at 6:39 PM, Alex Shi alex@intel.com wrote: This patch enabled the power aware consideration in load balance. As mentioned in the power aware scheduler proposal, Power aware scheduling has 2 assumptions: 1, race to idle is helpful for power saving 2, shrink tasks on less sched_groups will reduce power consumption The first assumption make performance policy take over scheduling when system busy. The second assumption make power aware scheduling try to move disperse tasks into fewer groups until that groups are full of tasks. This patch reuse lots of Suresh's power saving load balance code. Now the general enabling logical is: 1, Collect power aware scheduler statistics with performance load balance statistics collection. 2, if domain is eligible for power load balance do it and forget performance load balance, else do performance load balance. Has tried on my 2 sockets * 4 cores * HT NHM EP machine. and 2 sockets * 8 cores * HT SNB EP machine. In the following checking, when I is 2/4/8/16, all tasks are shrank to run on single core or single socket. $for ((i=0; i I; i++)) ; do while true; do : ; done done Checking the power consuming with a powermeter on the NHM EP. powersaving performance I = 2 148w160w I = 4 175w181w I = 8 207w224w I = 16 324w324w On a SNB laptop(4 cores *HT) powersaving performance I = 2 28w 35w I = 4 38w 52w I = 6 44w 54w I = 8 56w 56w On the SNB EP machine, when I = 16, power saved more than 100 Watts. Also tested the specjbb2005 with jrockit, kbuild, their peak performance has no clear change with powersaving policy on all machines. Just specjbb2005 with openjdk has about 2% drop on NHM EP machine with powersaving policy. This patch seems a bit long, but seems hard to split smaller. Signed-off-by: Alex Shi alex@intel.com -- Thanks
Re: [RFC PATCH 2/3] sched: power aware load balance,
Hi Alex, What I am concerned about in this patchset as Peter also mentioned in the previous discussion of your approach (https://lkml.org/lkml/2012/8/13/139) is that: 1.Using nr_running of two different sched groups to decide which one can be group_leader or group_min might not be be the right approach, as this might mislead us to think that a group running one task is less loaded than the group running three tasks although the former task is a cpu hogger. 2.Comparing the number of cpus with the number of tasks running in a sched group to decide if the group is underloaded or overloaded again faces the same issue.The tasks might be short running,not utilizing cpu much. I also feel before we introduce another side to the scheduler called 'power aware',why not try and see if the current scheduler itself can perform better? We have an opportunity in terms of PJT's patches which can help scheduler make more realistic decisions in load balance.Also since PJT's metric is a statistical one,I believe we could vary it to allow scheduler to do more rigorous or less rigorous power savings. It is true however that this approach will not try and evacuate nearly idle cpus over to nearly full cpus.That is definitely one of the benefits of your patch,in terms of power savings,but I believe your patch is not making use of the right metric to decide that. IMHO,the appraoch towards power aware scheduler should take the following steps: 1.Make use of PJT's per-entity-load tracking metric to allow scheduler to make more intelligent decisions in load balancing.Test the performance and power save numbers. 2.If the above shows some characteristic change in behaviour over the earlier scheduler,it should be either towards power save or towards performance.If found positive towards one of them, try varying the calculation of per-entity-load to see if it can lean towards the other behaviour.If it can,then there you go,you have a knob to change between policies right there! 3.If you don't get enough power savings with the above approach then add your patchset to evacuate nearly idle towards nearly busy groups,but by using PJT's metric to make the decision. What do you think? Regards Preeti U Murthy On Tue, Nov 6, 2012 at 6:39 PM, Alex Shi wrote: > This patch enabled the power aware consideration in load balance. > > As mentioned in the power aware scheduler proposal, Power aware > scheduling has 2 assumptions: > 1, race to idle is helpful for power saving > 2, shrink tasks on less sched_groups will reduce power consumption > > The first assumption make performance policy take over scheduling when > system busy. > The second assumption make power aware scheduling try to move > disperse tasks into fewer groups until that groups are full of tasks. > > This patch reuse lots of Suresh's power saving load balance code. > Now the general enabling logical is: > 1, Collect power aware scheduler statistics with performance load > balance statistics collection. > 2, if domain is eligible for power load balance do it and forget > performance load balance, else do performance load balance. > > Has tried on my 2 sockets * 4 cores * HT NHM EP machine. > and 2 sockets * 8 cores * HT SNB EP machine. > In the following checking, when I is 2/4/8/16, all tasks are > shrank to run on single core or single socket. > > $for ((i=0; i < I; i++)) ; do while true; do : ; done & done > > Checking the power consuming with a powermeter on the NHM EP. > powersaving performance > I = 2 148w160w > I = 4 175w181w > I = 8 207w224w > I = 16 324w324w > > On a SNB laptop(4 cores *HT) > powersaving performance > I = 2 28w 35w > I = 4 38w 52w > I = 6 44w 54w > I = 8 56w 56w > > On the SNB EP machine, when I = 16, power saved more than 100 Watts. > > Also tested the specjbb2005 with jrockit, kbuild, their peak performance > has no clear change with powersaving policy on all machines. Just > specjbb2005 with openjdk has about 2% drop on NHM EP machine with > powersaving policy. > > This patch seems a bit long, but seems hard to split smaller. > > Signed-off-by: Alex Shi -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
On Tue, 6 Nov 2012 21:09:58 +0800 Alex Shi wrote: > $for ((i=0; i < I; i++)) ; do while true; do : ; done & done > > Checking the power consuming with a powermeter on the NHM EP. > powersaving performance > I = 2 148w160w > I = 4 175w181w > I = 8 207w224w > I = 16 324w324w > > On a SNB laptop(4 cores *HT) > powersaving performance > I = 2 28w 35w > I = 4 38w 52w > I = 6 44w 54w > I = 8 56w 56w > > On the SNB EP machine, when I = 16, power saved more than 100 Watts. Confused. According to the above table, at I=16 the EP machine saved 0 watts. Typo in the data? Also, that's a pretty narrow test - it's doing fork and exec at very high frequency and things such as task placement decisions at process startup might be affecting the results. Also, the load will be quite kernel-intensive, as opposed to the more typical userspace-intensive loads. So, please run a broader set of tests so we can see the effects? -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
On Tue, 6 Nov 2012 21:09:58 +0800 Alex Shi alex@intel.com wrote: $for ((i=0; i I; i++)) ; do while true; do : ; done done Checking the power consuming with a powermeter on the NHM EP. powersaving performance I = 2 148w160w I = 4 175w181w I = 8 207w224w I = 16 324w324w On a SNB laptop(4 cores *HT) powersaving performance I = 2 28w 35w I = 4 38w 52w I = 6 44w 54w I = 8 56w 56w On the SNB EP machine, when I = 16, power saved more than 100 Watts. Confused. According to the above table, at I=16 the EP machine saved 0 watts. Typo in the data? Also, that's a pretty narrow test - it's doing fork and exec at very high frequency and things such as task placement decisions at process startup might be affecting the results. Also, the load will be quite kernel-intensive, as opposed to the more typical userspace-intensive loads. So, please run a broader set of tests so we can see the effects? -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [RFC PATCH 2/3] sched: power aware load balance,
Hi Alex, What I am concerned about in this patchset as Peter also mentioned in the previous discussion of your approach (https://lkml.org/lkml/2012/8/13/139) is that: 1.Using nr_running of two different sched groups to decide which one can be group_leader or group_min might not be be the right approach, as this might mislead us to think that a group running one task is less loaded than the group running three tasks although the former task is a cpu hogger. 2.Comparing the number of cpus with the number of tasks running in a sched group to decide if the group is underloaded or overloaded again faces the same issue.The tasks might be short running,not utilizing cpu much. I also feel before we introduce another side to the scheduler called 'power aware',why not try and see if the current scheduler itself can perform better? We have an opportunity in terms of PJT's patches which can help scheduler make more realistic decisions in load balance.Also since PJT's metric is a statistical one,I believe we could vary it to allow scheduler to do more rigorous or less rigorous power savings. It is true however that this approach will not try and evacuate nearly idle cpus over to nearly full cpus.That is definitely one of the benefits of your patch,in terms of power savings,but I believe your patch is not making use of the right metric to decide that. IMHO,the appraoch towards power aware scheduler should take the following steps: 1.Make use of PJT's per-entity-load tracking metric to allow scheduler to make more intelligent decisions in load balancing.Test the performance and power save numbers. 2.If the above shows some characteristic change in behaviour over the earlier scheduler,it should be either towards power save or towards performance.If found positive towards one of them, try varying the calculation of per-entity-load to see if it can lean towards the other behaviour.If it can,then there you go,you have a knob to change between policies right there! 3.If you don't get enough power savings with the above approach then add your patchset to evacuate nearly idle towards nearly busy groups,but by using PJT's metric to make the decision. What do you think? Regards Preeti U Murthy On Tue, Nov 6, 2012 at 6:39 PM, Alex Shi alex@intel.com wrote: This patch enabled the power aware consideration in load balance. As mentioned in the power aware scheduler proposal, Power aware scheduling has 2 assumptions: 1, race to idle is helpful for power saving 2, shrink tasks on less sched_groups will reduce power consumption The first assumption make performance policy take over scheduling when system busy. The second assumption make power aware scheduling try to move disperse tasks into fewer groups until that groups are full of tasks. This patch reuse lots of Suresh's power saving load balance code. Now the general enabling logical is: 1, Collect power aware scheduler statistics with performance load balance statistics collection. 2, if domain is eligible for power load balance do it and forget performance load balance, else do performance load balance. Has tried on my 2 sockets * 4 cores * HT NHM EP machine. and 2 sockets * 8 cores * HT SNB EP machine. In the following checking, when I is 2/4/8/16, all tasks are shrank to run on single core or single socket. $for ((i=0; i I; i++)) ; do while true; do : ; done done Checking the power consuming with a powermeter on the NHM EP. powersaving performance I = 2 148w160w I = 4 175w181w I = 8 207w224w I = 16 324w324w On a SNB laptop(4 cores *HT) powersaving performance I = 2 28w 35w I = 4 38w 52w I = 6 44w 54w I = 8 56w 56w On the SNB EP machine, when I = 16, power saved more than 100 Watts. Also tested the specjbb2005 with jrockit, kbuild, their peak performance has no clear change with powersaving policy on all machines. Just specjbb2005 with openjdk has about 2% drop on NHM EP machine with powersaving policy. This patch seems a bit long, but seems hard to split smaller. Signed-off-by: Alex Shi alex@intel.com -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/