Messages in this thread | | | Date | Fri, 06 Jul 2012 19:33:00 +0900 | From | Tomoki Sekiyama <> | Subject | Re: [RFC PATCH 00/18] KVM: x86: CPU isolation and direct interrupts handling by guests |
| |
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 <tomoki.sekiyama.qu@hitachi.com> Linux Technology Center Hitachi, Ltd., Yokohama Research Laboratory
| |