You could probably do it. I don't know how you would start to go about doing it with Windows though.
For instance how do you deal with the bootloader and Windows drive assignments? I suppose you could use BIOS harddrive emulation for USB devices.
I did it with Linux, but I wanted it to be portable. I had it setup to autoconfig the gui and drivers and harddrives and such each new computer I plugged into. I don't think that that would be possible with Windows.
But I suppose as long as you have it on one computer then that would be fine.
For performance it would be bad. Very bad. With Linux live-cdroms they use very high level of compression to get everything to fit on a cdrom. This helps because the compressed information can be read off the disk much faster then uncompressed information, the additional cpu overhead isn't that much in a modern machine.
I never tried a Windows-based live cdrom though, so I don't know how it compares. Does Windows 2003 support file system compression?
Also as far as real-world use goes the live linux cdroms use a ram-based file system layered over the read-only compressed file system. So actual use will be pretty fast as anything that is being read/written is all done in RAM.
On a USB key this sort of thing kills your performance.
So for server use it only makes sense to use a flash-based devices for a server were the server's file system remains relatively static.
That is for the majority of the time it's read-only. You start up the system, it starts up the services and those services stay running in RAM the entire time with very little swaping.
Then you still would have to have a real harddrive for data, swap file, and temporary files.
If you do that then a flash drive will easily outlast that harddrive. But if you keep the swap file and temporary files on the flash drive then it will wear out fairly quickly.
For instance a decent drive would have a read/write cycle lifetime of 100,000 times. So you could read/write to the entire drive a hundred thousand times. Also in addition they have load balancing systems so you don't wear out a paticular section of the drive faster then any other parts. It tries to spread it around. So with a 2 gigs of a 8gig flash drive used up with about 256 megs of data that gets writen/read to constantly then you maybe able to get around 2 million writes before you start to wear out things.
So say the flash drive was constantly busy. You have a real-world write peformance of about 5-6MB/s (which is pathetic, a typical harddrive has about what I expect would be 30-40MB/s in real usage.) (read speed is much faster)
So say you have that 256 meg workload, has 6 gigs of free space, and has a write speed of 6MB/s. It would take you 42 seconds to write that data out.
If I did the math correctly then that would allow about 3 years worth of life to that flash drive before it would wear out and you lost your data. That is under constant 24/7 use of writing over a aviable space of 6 gigs of spare room.
Now if you only had a gig worth of spare room then under constant writes it would wear out in about a year.
That's under 24/7 writing.
Now if you tailor your system to move temporary files to a ramdisk and eliminate the swap then you can easily have a flash drive that will outlast a harddrive if you use it for mostly read-only stuff.
If you have a lot of data you want to work with and have a read/write workload then sticking the OS on the flash drive, but having swap, temporary files, and work data on a real harddrive would work out fine.
but that's a lot of work.
Facebook
Twitter