Using a USB flash drive instead of a HD

[VirtualLarry]

I just came across an interesting page: http://www.ngine.de/index.jsp?pageid=4176
It purports to allow you to install WinXP onto a USB Mass Storage device, aka a flash drive or an external USB HD.

Given the extra space and power and of course cost associated with adding a HD to a dedicated cruncher, I thought that perhaps the possiblity of equipping said rig with a usb flash drive instead might be an interesting solution.

I plan on testing out that guide once I get back to my main rig.


I was also musing about the possibility of having an ATX power "Y" splitter, that would go from one PSU connector, to two mobo connectors. Such a thing does not exist yet, AFAIK, but it would seem feasable. All the PSU would see would be twice the load, which if the PSU was sized appropriately for the system(s) in question, would seem not to be a problem.

 

[NicColt]

very interesting, I wonder if this would work I havent tried it.

What if we booted with a system with Fedora 7 or Ubuntu Live and had the data on USB disk would that work ?

 

[The Reaper]

I heard that a flash drive can only read and write so many times before it wears out. Is this true? If so, will this work bonic?

 

[BadThad]

Wow! That's awesome! Thanks for the link.

As far as the reliability of flash drives....it's pretty bad. I've seen a lot of the bite the dust. Since most DC projects are normally CPU dependent, I don't see any particular advange other than a dedicated cruncher. In a crackrack situation, this could be quite handy and cheap....and probably cooler.

 

[The Borg]

I have actually had a flash drive mess up on me just running BOINC off of it - for a long time. That was not even the OS.

This is a brilliant idea, but I think the practicallities of the hardware lets us down.

As regards the PSU. I have re-wired 2 PSU's to run two machines each. The first one was two AT boards off one power supply. The other was an ATX PSU running an AT board and an ATX board. I have to start up the ATX board then both come alive. This is ok for normal running but not good for trouble shooting. Then you need to split the boards.

 

[Assimilator1]

Originally posted by: The Reaper
I heard that a flash drive can only read and write so many times before it wears out. Is this true? If so, will this work bonic?

Yes it is ,if ,BIG if, IIRC it's in the region of 10,000 times which in normal use isn't a problem ,but for the multiple accesses by an OS I don't know in addition to the DC project.Though some DC projects can alter the nu,mber of times it writes.

The Borg
How long did your flashdrive last running BOINC? (which project?)

 

[networkman]

Thanks for the link - it's interesting enough to me to warrant at least giving it a try.

 

[The Borg]

Cannot remmeber which project, it was a while back. Time wise, I think it was over a weekend. Very short. Noticably short enough to realise what the reason was - too much access to the flash drive.

 

[Alyx]

I'm amazed that they will only write a limited number of times, I wonder what the reason for this is. I'd have thought they'd work untill the USB connector metal wore down to much. I've got one that went through the washer and works like a charm and another that I accidentally bent nearly in half right where the plastic ends and the connector starts when rotating a machine around, it looks ghetto, I almost feel it needs duck tape to hide the broken plastic, but it still works.

 

[Assimilator1]

Originally posted by: The Borg
Cannot remmeber which project, it was a while back. Time wise, I think it was over a weekend. Very short. Noticably short enough to realise what the reason was - too much access to the flash drive.

A? ,that doesn't make sense ,you said up above that I have actually had a flash drive mess up on me just running BOINC off of it - for a long time. & here you say it was over a weekend ,which is it?

 

[The Borg]

Oops, sorry for the confusion. Memory a bit fuzzy. I think it was about a weekend's worth of running. What I meant by 'long time' is that I was actually running BOINC off the flash drive longer than just downloading work for putting back on a non-internet connected machine.

 

[VirtualLarry]

I can't imagine that running an OS could be worse than using a flash drive for ReadyBoost purposes, could it be?
There's an article at DansData, about using a flash drive for the pagefile (no seeking!), and he makes some calculations, that essentially point out that the longevity of the flash drive is also related to the size and overall number of sectors, assuming that writes are spread out evenly among those sectors. So using a suitably large-sized drive, should give enough longevity. I guess it remains to be seen what that size is.

I haven't tried the guide yet, still been messing around with my RAID array (softraid is a PITA!), keeps dropping drives for no good reason.

 

[keeleysam]

If you're going to do that, use BOINCpe, as you won't be constantly paging to the flash drive.

 

[Philippart]

that tutorial seems somewhat complicated if you only want to test it, a much easier tutorial is here with pendrive linux.

 

[Assimilator1]

Originally posted by: The Borg
Oops, sorry for the confusion. Memory a bit fuzzy. I think it was about a weekend's worth of running. What I meant by 'long time' is that I was actually running BOINC off the flash drive longer than just downloading work for putting back on a non-internet connected machine.

Ah I see ,fair enough ,though in which case I very much doubt that it's BOINC that killed your flash drive in such a short time ,I think that was just a coincedence.
As far as I remember the number of writes you'd get from a DC project would take many many months (maybe a few years or more?) to kill a flash drive.