[Tfug] OT: Disk testing

[Ronald Sutherland]

You may be right on most accounts, testing HD's is not the sort of thing 
I do, I do get payed to develop test for power conversion products, but 
its not the same. In general I don't like to invent, its usually a wast 
of my time, and it always gets undone in a few years when the setup has 
to go to Mexico or China. I like things with a well defined interface, 
both hardware and software, I've even made up my own term "Modular 
Components Off The Shelf". Its Maintained, invented and extended by 
not-me, but it should be open if needed. If the ATA interface is simple, 
thats all the more reason to use it. Adding layers like USB or Ethernet, 
have caused me some problems with other data buses, and I have limited 
kernel or driver software skills, although I have done a lot with 8 bit 
MCU's without a proper OS. So the idea of an MCU sounds good, but 
basically that is what CORAID is doing with ATA over Ethernet, which 
would also seem elegant.

http://www.coraid.com/

The problem with the Coraid product is that it will not stand up to much 
abuse. If speed is not an issue, then I have no idea what "comprehensive 
surface analysis" is, anyway USB is clean at least for wiring, however a 
mobile rack type thing can help when you are dealing with a lot of 
parallel control lines. The wiring mess is in an enclosed fixture and 
the device under test is put in a carrier, then loaded into the fixture, 
its a basic method used to test many products. PC's do eat a lot of 
space, and thats a real problem, As far as power, a PC is less wasteful 
over all. Every power conversion device eats some bias power (2% or 
more) and smaller supply's are the most, with high bias (~10%) and are 
less efficient. So a single PC supply (say 600W) running everything is 
about the most efficient method power wise. Many PC's have a power 
supply that is a real POS and need replaced with a good one that 
includes PFC. The power supply will only use what is needed, plus its 
bias power, and efficiency factor.

Bexley Hall wrote:
> Hi, Ronald,
>
> --- Ronald Sutherland <rsutherland at epccs.com> wrote:
>
>   
>> I'm thinking that USB will not scale past a few
>>     
> HD's,
>
> Yes, but adding more PC's uses up more bench space,
> power, etc.  It's just not a very elegant solution  :<
> And, one could always add more USB i/f's until you
> max out the PCI bus.
>
> (The "ideal" solution is a small microcontroller
> attached to each drive to just push bytes in/out
> that one drive and report results to a host that
> presents the user interface -- but, I don't want
> to have to layout a board for this...)
>
>   
>> I guess I'd personally be thinking along the 
>> lines of multiple PC's. With a test 
>> program that runs after startup in each. I'd
>>     
>
> Yes, if you ignore the cost of space and the
> wasted power running a whole pc instead of
> *just* a disk drive...  :<
>
>   
>> probably also get some 
>> cheep IDE cards to add IDE channels, I've seen some
>> that add 2 (or 4) 
>> IDE channels from HighPoint and others, and if I
>> remember they claim to work in Linux.
>>     
>
> An IDE card is little more than a 16 bit buffer
> and a connector.  Making one work in any OS is
> a cakewalk.  :>
>
>   
>> After adding 4 IDE channels, one
>> master drive per 
>> channel, that should give 6 master IDE's total, keep
>>     
>
> But now you have 5 disk drives dangling on 16"
> ribbon cables sprawled on a desktop.  I have a hard
> enough time getting two disks to dress nicely in a
> single case -- let alone one inside and 5 outside!
> (you don't want to ount the drives inside unless you
> can do so without fasteners... which means you
> now have to look for a case that will accommodate
> this, etc.)
>
>   
>> one for system and 
>> use the rest for testing. Since I've never tried
>> this sort of thing, I 
>> wonder if its possible to kill the power to an IDE
>> master only and 
>> remove it while leaving the other masters running.
>>     
>
> You can do this -- assuming no slave attached.
> But, it is not "good form".  You are just waiting
> for something to die, somewhere.  Putting a
> mechanical switch (DPST) in line with the power
> connection can help.  I just feel it is easier to
> power off an external USB drive enclosure (which
> is already *expecting* to be used in this fashion)
> and then unplugging the cables.
>
>   
>> Also I think that the 
>> PCI data bus will saturate around 5 to 6 HD's so
>> more than that per 
>> system may not be faster, if I'm wrong about that
>> then fill more PCI slots with IDE cards.
>>     
>
> It depends on how you are exercising the drives
> and the characteristics of the drives.  E.g., if
> you deliberately do random seeks, then you invalidate
> any internal cache benefits for the drives and are
> limited by access times, not bandwidth.
>
> Also, I think some newer drives can be told to
> format themselves thereby cutting down on bus
> usage, etc.
>
> (sigh)  It may prove easier to just hack together
> a small microcontroller and stuff one on each
> drive.  This would eliminate all of the cabling
> issues, bus bandwidth limitations, power supply,
> etc.  I'll have to see what I can find for a single
> chip solution as that might let me get away with
> a little 2"x2" PCB -- perhaps just plug it
> right onto the back of the drive!  :>
>
> Thanks!
> --don
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around 
> http://mail.yahoo.com 
>
> _______________________________________________
> Tucson Free Unix Group - tfug at tfug.org
> Subscription Options:
> http://www.tfug.org/mailman/listinfo/tfug_tfug.org
>
>