Which Workstation host O/S has the lowest memory requirement?

[UlyssesOfEpirus]

Some time ago it was possible to install Workstation in damnsmalllinux, after following some instructions to rebuild the kernel. It was amazingly low in its memory requirement, as low as 32mb or so. If only the usb support worked on my hardware. It would have made the perfect host for vmware.

What other operating systems can be hosts for vmware workstation while keeping very little memory for themselves?

[continuum]

are you looking for performance or is this for using WS from a LiveCD or on machine with little RAM ?

In my experience for best overall performance I prefer to load a lean host OS into RAM ... obviously you will lose some RAM for the VMs but this is more then compensated by the way faster host.

If I compare a LinuxLiveCD with WS to the 2003 BartPE LiveCDs I make - they load into RAM completely - the BartPE outruns all Linux LiveCDs with Workstation I ever made or seen




_________________________

VMX-parameters- WS FAQ -[ MOAcd|http://sanbarrow.com/moa241.html] - VMDK-Handbook

[UlyssesOfEpirus]

It's for a machine that does not have enough memory, even the 2gb it comes with is not enough for me because the host O/S seems to be so bloated that only 3 VM's of 400mb each are allowed in memory at a time (swapping is not allowed for VM's, for performance, see config.ini below).

prefvmx.minVmMemPct = "100"

sched.mem.pshare.enable = "FALSE"

prefvmx.useRecommendedLockedMemSize = "TRUE"

MemTrimRate = "0"

MemAllowAutoScaleDown = "FALSE"

I do not understand why a liveCD would make VM's run so much faster than a normal installation. On windows xp, Process Explorer shows the cpu as almost idle when the VM's are idle. Therefore all processing power is available to the VM's, therefore it shouldn't make noticeable difference if the host o/s is entirely in memory. So how do you explain your observation that VM's are much faster on a liveCD of windows 2003, than on a normal installation of the same O/S?

[wila]

Hello,

PMJI, is the live CD expert here, but let me answer this one.

So how do you explain your observation that VM's are much faster on a liveCD of windows 2003, than on a normal installation of the same O/S

A liveCD like BartPE that Ulli talks about is a stripped down version of windows 2003, it doesn't run all the same services and background processes that a default windows 2003 setup would run.

I'm not sure I'd agree with Ulli's statement on Windows liveCDs always requiring less memory, but maybe he means that in regards to running VMware guests, as in that case it might be true.

--
Wil
_____________________________________________________
VI-Toolkit & scripts wiki at http://www.vi-toolkit.com

Contributing author at blog www.planetvm.net

Twitter: @wilva

[continuum]

I do not understand why a liveCD would make VM's run so much faster than a normal installation

a regular Windows 2003 may have 30 processes running - a 2003 LiveCD only uses one third of them.

Additionally having all the Windows and VMware Workstation files loaded into RAM gives another speed boost.

_________________________

VMX-parameters- WS FAQ -[ MOAcd|http://sanbarrow.com/moa241.html] - VMDK-Handbook

[UlyssesOfEpirus]

So can I expect a performance increase of 20% in my VM's if I switch to the windows 2003 liveCD, compared to my current windows XP SP2 setup?

Has anyone benchmarked VM's under that liveCD, and compared with the same under windows XP? Judging from the output of Process Explorer (1-3% of cpu time going to XP processes), reducing the number of xp processes to 0 would make VM's run 1-3% faster. Is the math correct here, or that's not how things work?

I honestly have doubts about the correctness of the above math because I've seen weird benchmark results with video encoding in VM's. Try avidemux on various linux distros in a VM (this has its own standalone builtin codec). Even though the system monitor reports the CPU is almost idle in all distros when not encoding, the performance varies dramatically.