1 person found this helpful
1) Will this hardware and/or OS work with VMware server?
It should work just fine.
2) Does the X2 contain virtualization facilities (Pacifica? SVM?) and if so does VMware server take advantage of them (I have searched google, unsuccessfully!)
VMware Server does not use any of the hardware virtualization at this point in time - their software version has better performance. The only place they DO use it is on Intel platforms, to support 64-bit guests. See the tech. paper here for more info http://www.vmware.com/vmtn/resources/528
Cool - thanks for the information.
Do you know if it's possible to set the logical processor affinity on a per-VM basis? So I could have two VMs, each on a separate core to maximise the efficiency of the box.
Yes, you can do this if you really want to. Most of the time, the host OS's process scheduler already does a better job than most of us can do manualy though.
If you ever get a \*stable* host OS let me know. Because I had a hard time dealing with the ACPI in AMD64 Linux kernels..
I used to run Gentoo 2006.1 onver a AMD64 X2. It was like a firecracker with VMware Server.... I had two system crash with it
I use now i686 Debian based distro for my AMD64 X2 server.. The ACPI works without problems.
I do care for host OS optimisations. Stability is misson-critital in my virtual machines.
I suggest you benchmark and stress test your vm before giving it any dedicated purpose.
On second tought, your Ubuntu is Debian based. It should be more stable.
If you don't get any ACPI issues over your AMD64 X2 please let me know!! I would be pleased to run a 64bit OS for my host..
None of the SuSE versions that use have ACPI problems (or any problems for that matter) - 32 or 64 bit versions. The one exception is that the SuSE 10.2 installation inside a VM sometimes fails with an ACPI error on one of my P3-based devel servers. That's easy enough to correct, though - I force ACPI to be disabled for the install and once the install is done the kernel deals with ACPI correctly.
Yes, you can do this if you really want to. Most of
the time, the host OS's process scheduler already
does a better job than most of us can do manualy
This is not true in AMD multi-CPU environments.
When a task gets swapped between CPUs, memory blocks most often end up being created on the various CPUs with the CPUs having to access different memory blocks through one another.
On AMD servers, it is much better to assign CPU affinity and confine each VM to a single physical CPU.
Yes, memory can still end up being created on adjoining CPUs, but this is not as often as when tasks get swapped between physical CPUs.
I did some benchmarks to prove that.