[Server-devel] Western Digital hard disk data

[James Cameron]

Thanks, that all sounds reasonable.

It is not practical to make clear recommendations from these tests,
because of the number of dimensions in the phase space that have not
been explored:

- how other external drives of the same model draw power and respond
  to the bus voltage, (manufacturing variance within a batch, and
  changes to components used in manufacturing for which tracking
  information is not available to you),

- how other external drives of other models draw power and respond to
  the bus voltage,

- how other XO-1.5, XO-1.75, or XO-4 laptops respond to the power
  draw, (our manufacturing variable and changes to components used in
  our manufacturing),

- how the environment temperature affects the results,

- how the aging of the hard drive mechanism changes the startup power,

- how the aging of the laptop changes the voltage droop.

While a powered USB hub may appear to be a solution, such a hub could
suffer from the same kinds of behaviour, and variance.  It also adds
more dimensions to the phase space.  It also adds one more source of
power drain in a remote system.

A reasonable qualification test could be written.  Given an XO being
deployed as XSCE, with a sample hard drive, and a USB ethernet device,
adding a third USB load and test ten power cycles.  You might also use
a variable current draw device to find the power headroom in the
system before failure to spin up.

You can construct a variable current draw device from a switching
voltage regulator feeding a variable resistor.  Smoke might be
released if the power exceeds the maximum rating of the resistor or
the regulator.

There's probably something else I haven't thought of.  Consider this a
hasty response.

-- 
James Cameron
http://quozl.linux.org.au/