Jan, Nicolas, Thank you for your help:
> Third option may be even more interesting (provided you really want to > use Windows for the UI): Run Xenomai on the host and combine it with KVM > to host another OS. This works fine with latest ipipe for 3.8 when RT > tasks and the QEMU process (that uses KVM) are on separate CPUs, I just > heard of issues (couldn't reproduce yet) when they share a CPU. > If I am not wrong, KVM runs on top of Linux, and then Windows on top of it. Isn't it too much overhead? Isn't it better to use a Type 1 virtualization? When you say separate CPUs, do you mean separate cores? > Virtualizing your RT workload won't make things faster, naturally. > Depending on the hypervisor architecture and hardware capabilities, it > may not be measurable. > So, do you mean that, even though the system may be slower, the real-time response will run smoothly? > But I would refrain from such stacking unless there is a real need > (which I do not see for your setup). > My plan for the mid term is to run the user interface (HMI) and the real-time applications in one board. Meanwhile, I need to update my HMI and I want to use a complete Linux distribution (one with a desktop manager) with Xenomai. However, I may finally be forced to use Windows (3rd party programs) and, when porting my system to one board, I would need some virtualization tools. Yesterday, I wrote another e-mail asking for opinions about a Linux distribution with Xenomai, or about any experience with the Yocto project and Xenomai. Has anyone got any information regarding this? > In Windows world, use RTX : http://www.intervalzero.com/products/rtx-rtx64/overview/ > In fact, if I change from Xenomai Linux to Windows, first I will remove all the real-time applications from that board and run them in the other one. Anyway, it's an interesting link. Thanks again, Ata 2013/8/29 Jan Kiszka <[email protected]> > On 2013-08-29 13:06, Asier Tamayo Arbide wrote: > > Hello all, > > > > My embedded system has two boards running Linux: one runs the user > > interface and the other one the critical real-time tasks. Both boards are > > running Xenomai on an Atom N270 chipset. > > > > Now, I am planning to compact my system in just one board and have two > > options. The first one is to execute everything using Xenomai and the > > second one is to port my user interface to Windows and to run both > Windows > > and Xenomai on top of a hypervisor. My e-mail regards this second option. > > > > Third option may be even more interesting (provided you really want to > use Windows for the UI): Run Xenomai on the host and combine it with KVM > to host another OS. This works fine with latest ipipe for 3.8 when RT > tasks and the QEMU process (that uses KVM) are on separate CPUs, I just > heard of issues (couldn't reproduce yet) when they share a CPU. > > > I would like to know if anyone has any experience with this. Is the > > real-time response affected in any way when using a hypervisor? My > current > > kernel has been compiled using the ELinOS Embedded Linux distribution. > Do I > > have to patch it? > > > > Virtualizing your RT workload won't make things faster, naturally. > Depending on the hypervisor architecture and hardware capabilities, it > may not be measurable. But I would refrain from such stacking unless > there is a real need (which I do not see for your setup). > > > I am studying Xen, VmWare, KVM and Enea. Do you recommend any of them? > The > > board I’ll use in the new (compact) system will support virtualization > > extensions (Intel VT). > > See above. > > Jan > > -- > Siemens AG, Corporate Technology, CT RTC ITP SES-DE > Corporate Competence Center Embedded Linux > _______________________________________________ Xenomai mailing list [email protected] http://www.xenomai.org/mailman/listinfo/xenomai
