From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tomoki Sekiyama Subject: Re: [RFC PATCH 00/18] KVM: x86: CPU isolation and direct interrupts handling by guests Date: Fri, 06 Jul 2012 19:33:00 +0900 Message-ID: <4FF6BEDC.8040207@hitachi.com> References: <20120628060719.19298.43879.stgit@localhost.localdomain> <4FEC8D31.3070406@redhat.com> <4FEC93BD.2070809@siemens.com> <4FEC95BC.200@redhat.com> <4FED7495.30707@hitachi.com> <4FEDC22F.9070406@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: jan.kiszka@siemens.com, kvm@vger.kernel.org, linux-kernel@vger.kernel.org, x86@kernel.org, yrl.pp-manager.tt@hitachi.com To: avi@redhat.com Return-path: In-Reply-To: <4FEDC22F.9070406@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org Hi, On 2012/06/29 23:56, Avi Kivity wrote: >> On 2012/06/29 2:34, Avi Kivity wrote: >>> On 06/28/2012 08:26 PM, Jan Kiszka wrote: >>>>> This is both impressive and scary. What is the target scenario here? >>>>> Partitioning? I don't see this working for generic consolidation. >>>> >>>> From my POV, partitioning - including hard realtime partitions - would >>>> provide some use cases. >> >> Exactly this is for partitioning that requires bare-metal performance >> with low latency and realtime. > > It's hard for me to evaluate how large that segment is. Since the > patchset is so intrusive, it needs a large potential user set to > justify, or a large reduction in complexity, or both. Low latency or realtime is often required on high-end systems like trading, automated control, HPC and so on, or for multimedias. Those who want to run MRG as a guest, or to fully utilize high-speed NIC are also worth using this. And not all of such applications does not use up every CPU, so partitioning is becoming reasonable as a number of cores in a server is increasing. Anyway, I will try to make the patch as simple as possible. >> I think it is also useful for workload >> like HPC with MPI, that is CPU intensive and that needs low latency. > > I keep hearing about people virtualizing these types of workloads, but I > haven't yet understood why. One reason is ease of deployment of applications to nodes. Especially in IaaS environment like Amazon EC2 Cluster Compute Instances, virtualization is often introduced as a simple way to move applications around flexibly among nodes shared by many users. Thanks, -- Tomoki Sekiyama Linux Technology Center Hitachi, Ltd., Yokohama Research Laboratory