Re: [openstack-dev] [nova] Configure overcommit policy
In fact, there is a blueprint which would enable supporting this scenario without partitioning -- https://blueprints.launchpad.net/nova/+spec/cpu-entitlement The idea is to annotate flavors with CPU allocation guarantees, and enable differentiation between instances, potentially running on the same host. The implementation is augmenting the CoreFilter code to factor in the differentiation. Hopefully this will be out for review soon. Regards, Alex From: John Garbutt j...@johngarbutt.com To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date: 14/11/2013 04:57 PM Subject:Re: [openstack-dev] [nova] Configure overcommit policy On 13 November 2013 14:51, Khanh-Toan Tran khanh-toan.t...@cloudwatt.com wrote: Well, I don't know what John means by modify the over-commit calculation in the scheduler, so I cannot comment. I was talking about this code: https://github.com/openstack/nova/blob/master/nova/scheduler/filters/core_filter.py#L64 But I am not sure thats what you want. The idea of choosing free host for Hadoop on the fly is rather complicated and contains several operations, namely: (1) assuring the host never get pass 100% CPU load; (2) identifying a host that already has a Hadoop VM running on it, or already 100% CPU commitment; (3) releasing the host from 100% CPU commitment once the Hadoop VM stops; (4) possibly avoiding other applications to use the host (to economy the host resource). - You'll need (1) because otherwise your Hadoop VM would come short of resources after the host gets overloaded. - You'll need (2) because you don't want to restrict a new host while one of your 100% CPU commited hosts still has free resources. - You'll need (3) because otherwise you host would be forerever restricted, and that is no longer on the fly. - You'll may need (4) because otherwise it'd be a waste of resources. The problem of changing CPU overcommit on the fly is that when your Hadoop VM is still running, someone else can add another VM in the same host with a higher CPU overcommit (e.g. 200%), (violating (1) ) thus effecting your Hadoop VM also. The idea of putting the host in the aggregate can give you (1) and (2). (4) is done by AggregateInstanceExtraSpecsFilter. However, it does not give you (3); which can be done with pCloud. Step 1: use flavors so nova can tell between the two workloads, and configure them differently Step 2: find capacity for your workload given your current cloud usage At the moment, most of our solutions involve reserving bits of your cloud capacity for different workloads, generally using host aggregates. The issue with claiming back capacity from other workloads is a bit tricker. The issue is I don't think you have defined where you get that capacity back from? Maybe you want to look at giving some workloads a higher priority over the constrained CPU resources? But you will probably starve the little people out at random, which seems bad. Maybe you want to have a concept of spot instances where they can use your spare capacity until you need it, and you can just kill them? But maybe I am miss understanding your use case, its not totally clear to me. John ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] Configure overcommit policy
Step 1: use flavors so nova can tell between the two workloads, and configure them differently Step 2: find capacity for your workload given your current cloud usage At the moment, most of our solutions involve reserving bits of your cloud capacity for different workloads, generally using host aggregates. The issue with claiming back capacity from other workloads is a bit tricker. The issue is I don't think you have defined where you get that capacity back from? Maybe you want to look at giving some workloads a higher priority over the constrained CPU resources? But you will probably starve the little people out at random, which seems bad. Maybe you want to have a concept of spot instances where they can use your spare capacity until you need it, and you can just kill them? But maybe I am miss understanding your use case, its not totally clear to me. Yes currently we can only reserve some hosts for particular workloads. But «reservation» is done by admins operation, not «on-demand» as I understand. Anyway, its just some speculations from what I think of Alexander usecase. Or maybe I misunderstand Alexander ? It is interesting to see the development of the CPU entitlement blueprint that Alex mentioned. It was registered in Jan 2013. Any idea whether it is still going on? De : Alex Glikson [mailto:glik...@il.ibm.com] Envoyé : jeudi 14 novembre 2013 16:13 À : OpenStack Development Mailing List (not for usage questions) Objet : Re: [openstack-dev] [nova] Configure overcommit policy In fact, there is a blueprint which would enable supporting this scenario without partitioning -- https://blueprints.launchpad.net/nova/+spec/cpu-entitlement The idea is to annotate flavors with CPU allocation guarantees, and enable differentiation between instances, potentially running on the same host. The implementation is augmenting the CoreFilter code to factor in the differentiation. Hopefully this will be out for review soon. Regards, Alex From:John Garbutt j...@johngarbutt.com To:OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date:14/11/2013 04:57 PM Subject:Re: [openstack-dev] [nova] Configure overcommit policy _ On 13 November 2013 14:51, Khanh-Toan Tran khanh-toan.t...@cloudwatt.com wrote: Well, I don't know what John means by modify the over-commit calculation in the scheduler, so I cannot comment. I was talking about this code: https://github.com/openstack/nova/blob/master/nova/scheduler/filters/core _filter.py#L64 https://github.com/openstack/nova/blob/master/nova/scheduler/filters/core_ filter.py#L64 But I am not sure thats what you want. The idea of choosing free host for Hadoop on the fly is rather complicated and contains several operations, namely: (1) assuring the host never get pass 100% CPU load; (2) identifying a host that already has a Hadoop VM running on it, or already 100% CPU commitment; (3) releasing the host from 100% CPU commitment once the Hadoop VM stops; (4) possibly avoiding other applications to use the host (to economy the host resource). - You'll need (1) because otherwise your Hadoop VM would come short of resources after the host gets overloaded. - You'll need (2) because you don't want to restrict a new host while one of your 100% CPU commited hosts still has free resources. - You'll need (3) because otherwise you host would be forerever restricted, and that is no longer on the fly. - You'll may need (4) because otherwise it'd be a waste of resources. The problem of changing CPU overcommit on the fly is that when your Hadoop VM is still running, someone else can add another VM in the same host with a higher CPU overcommit (e.g. 200%), (violating (1) ) thus effecting your Hadoop VM also. The idea of putting the host in the aggregate can give you (1) and (2). (4) is done by AggregateInstanceExtraSpecsFilter. However, it does not give you (3); which can be done with pCloud. Step 1: use flavors so nova can tell between the two workloads, and configure them differently Step 2: find capacity for your workload given your current cloud usage At the moment, most of our solutions involve reserving bits of your cloud capacity for different workloads, generally using host aggregates. The issue with claiming back capacity from other workloads is a bit tricker. The issue is I don't think you have defined where you get that capacity back from? Maybe you want to look at giving some workloads a higher priority over the constrained CPU resources? But you will probably starve the little people out at random, which seems bad. Maybe you want to have a concept of spot instances where they can use your spare capacity until you need it, and you can just kill them? But maybe I am miss understanding your use case, its not totally clear to me. John ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http
Re: [openstack-dev] [nova] Configure overcommit policy
Khanh-Toan Tran khanh-toan.t...@cloudwatt.com wrote on 14/11/2013 06:27:39 PM: It is interesting to see the development of the CPU entitlement blueprint that Alex mentioned. It was registered in Jan 2013. Any idea whether it is still going on? Yes. I hope we will be able to rebase and submit for review soon. Regards, Alex De : Alex Glikson [mailto:glik...@il.ibm.com] Envoyé : jeudi 14 novembre 2013 16:13 À : OpenStack Development Mailing List (not for usage questions) Objet : Re: [openstack-dev] [nova] Configure overcommit policy In fact, there is a blueprint which would enable supporting this scenario without partitioning -- https://blueprints.launchpad.net/ nova/+spec/cpu-entitlement The idea is to annotate flavors with CPU allocation guarantees, and enable differentiation between instances, potentially running on thesame host. The implementation is augmenting the CoreFilter code to factor in the differentiation. Hopefully this will be out for review soon. Regards, Alex From:John Garbutt j...@johngarbutt.com To:OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date:14/11/2013 04:57 PM Subject:Re: [openstack-dev] [nova] Configure overcommit policy On 13 November 2013 14:51, Khanh-Toan Tran khanh-toan.t...@cloudwatt.com wrote: Well, I don't know what John means by modify the over-commit calculation in the scheduler, so I cannot comment. I was talking about this code: https://github.com/openstack/nova/blob/master/nova/scheduler/ filters/core_filter.py#L64 But I am not sure thats what you want. The idea of choosing free host for Hadoop on the fly is rather complicated and contains several operations, namely: (1) assuring the host never get pass 100% CPU load; (2) identifying a host that already has a Hadoop VM running on it, or already 100% CPU commitment; (3) releasing the host from 100% CPU commitment once the Hadoop VM stops; (4) possibly avoiding other applications to use the host (to economy the host resource). - You'll need (1) because otherwise your Hadoop VM would come short of resources after the host gets overloaded. - You'll need (2) because you don't want to restrict a new host while one of your 100% CPU commited hosts still has free resources. - You'll need (3) because otherwise you host would be forerever restricted, and that is no longer on the fly. - You'll may need (4) because otherwise it'd be a waste of resources. The problem of changing CPU overcommit on the fly is that when your Hadoop VM is still running, someone else can add another VM in the same host with a higher CPU overcommit (e.g. 200%), (violating (1) ) thus effecting your Hadoop VM also. The idea of putting the host in the aggregate can give you (1) and (2). (4) is done by AggregateInstanceExtraSpecsFilter. However, it does not give you (3); which can be done with pCloud. Step 1: use flavors so nova can tell between the two workloads, and configure them differently Step 2: find capacity for your workload given your current cloud usage At the moment, most of our solutions involve reserving bits of your cloud capacity for different workloads, generally using host aggregates. The issue with claiming back capacity from other workloads is a bit tricker. The issue is I don't think you have defined where you get that capacity back from? Maybe you want to look at giving some workloads a higher priority over the constrained CPU resources? But you will probably starve the little people out at random, which seems bad. Maybe you want to have a concept of spot instances where they can use your spare capacity until you need it, and you can just kill them? But maybe I am miss understanding your use case, its not totally clear to me. John ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev Aucun virus trouvé dans ce message. Analyse effectuée par AVG - www.avg.fr Version: 2014.0.4158 / Base de données virale: 3629/6834 - Date: 13/11/2013 ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] Configure overcommit policy
On 11 November 2013 12:04, Alexander Kuznetsov akuznet...@mirantis.com wrote: Hi all, While studying Hadoop performance in a virtual environment, I found an interesting problem with Nova scheduling. In OpenStack cluster, we have overcommit policy, allowing to put on one compute more vms than resources available for them. While it might be suitable for general types of workload, this is definitely not the case for Hadoop clusters, which usually consume 100% of system resources. Is there any way to tell Nova to schedule specific instances (the ones which consume 100% of system resources) without overcommitting resources on compute node? You could have a flavor with no-overcommit extra spec, and modify the over-commit calculation in the scheduler on that case, but I don't remember seeing that in there. John ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] Configure overcommit policy
You can consider having a separate host aggregate for Hadoop, and use a combination of AggregateInstanceExtraSpecFilter (with a special flavor mapped to this host aggregate) and AggregateCoreFilter (overriding cpu_allocation_ratio for this host aggregate to be 1). Regards, Alex From: John Garbutt j...@johngarbutt.com To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date: 12/11/2013 04:41 PM Subject:Re: [openstack-dev] [nova] Configure overcommit policy On 11 November 2013 12:04, Alexander Kuznetsov akuznet...@mirantis.com wrote: Hi all, While studying Hadoop performance in a virtual environment, I found an interesting problem with Nova scheduling. In OpenStack cluster, we have overcommit policy, allowing to put on one compute more vms than resources available for them. While it might be suitable for general types of workload, this is definitely not the case for Hadoop clusters, which usually consume 100% of system resources. Is there any way to tell Nova to schedule specific instances (the ones which consume 100% of system resources) without overcommitting resources on compute node? You could have a flavor with no-overcommit extra spec, and modify the over-commit calculation in the scheduler on that case, but I don't remember seeing that in there. John ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] Configure overcommit policy
FYI, by default Openstack overcommit CPU 1:16, meaning it can host 16 times number of cores it possesses. As mentioned Alex, you can change it by enabling AggregateCoreFilter in nova.conf: scheduler_default_filters = list of your filters, adding AggregateCoreFilter here and modifying the overcommit ratio by adding: cpu_allocation_ratio=1.0 Just a suggestion, think of isolating the host for the tenant that uses Hadoop so that it will not serve other applications. You have several filters at your disposal: AggregateInstanceExtraSpecsFilter IsolatedHostsFilter AggregateMultiTenancyIsolation Best regards, Toan - Original Message - From: Alex Glikson glik...@il.ibm.com To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org Sent: Tuesday, November 12, 2013 3:54:02 PM Subject: Re: [openstack-dev] [nova] Configure overcommit policy You can consider having a separate host aggregate for Hadoop, and use a combination of AggregateInstanceExtraSpecFilter (with a special flavor mapped to this host aggregate) and AggregateCoreFilter (overriding cpu_allocation_ratio for this host aggregate to be 1). Regards, Alex From: John Garbutt j...@johngarbutt.com To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date: 12/11/2013 04:41 PM Subject: Re: [openstack-dev] [nova] Configure overcommit policy On 11 November 2013 12:04, Alexander Kuznetsov akuznet...@mirantis.com wrote: Hi all, While studying Hadoop performance in a virtual environment, I found an interesting problem with Nova scheduling. In OpenStack cluster, we have overcommit policy, allowing to put on one compute more vms than resources available for them. While it might be suitable for general types of workload, this is definitely not the case for Hadoop clusters, which usually consume 100% of system resources. Is there any way to tell Nova to schedule specific instances (the ones which consume 100% of system resources) without overcommitting resources on compute node? You could have a flavor with no-overcommit extra spec, and modify the over-commit calculation in the scheduler on that case, but I don't remember seeing that in there. John ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
