Hardware Reference
In-Depth Information
The desktop system was from about 48% to 55% faster than the laptop on USB transfers, yet only 1%
to 3% faster on FireWire transfers. In other words, the FireWire transfer times were virtually
identical
on both the desktop and laptop systems, even though the systems varied greatly in speed.
This is understandable because FireWire uses a peer-to-peer connection design, which does not use
the PC to manage the interface. On the other hand, USB is a PC-centric design using the PC as the host
controller. In other words, because of the internal design differences in USB and FireWire, the PC
processor and overall PC system performance have a much greater effect on USB than on FireWire.
Another interesting point: The method used to copy the file (
COPY
versus
XCOPY
in this case)
actually made
more
of a difference than the interface or the system speed. This brings up many more
questions: Would there have been other differences had the drive been formatted with NTFS instead
of FAT32? What if I had copied many small files instead of one large file? What if I had used a
different operating system (OS)? As you can see, in any benchmark or comparison, you must consider
a huge number of variables—and there are no simple answers to ambiguous questions!
Hot-Plugging (and Unplugging)
Because external USB and 1394 (FireWire) devices can be hot-plugged, you must be careful in some
cases when unplugging things, especially storage devices. To help prevent data loss with external
drives and storage devices, when disconnecting them you should click the Safely Remove Hardware
icon in the Windows taskbar notification area (often called the
system tray
). Then select the device to
be removed and wait for the system to indicate that the device can be safely removed before you
unplug it.
But consider this: I've unplugged devices without “safely removing” them many times without
problems, so is it really necessary to go to the trouble? To answer that question properly, I'll have to
explain more about how caching works, both in Windows and in the hardware.
Windows includes a write cache in the OS (using system RAM) Microsoft likes to call
lazy write
.
The OS-based lazy write function has been around for a long time. In fact, it has been enabled by
default for nonremovable drives (internal ATA hard drives, for example) in all versions of Windows
since version 3.1 (as smartdrv.exe), which dates from 1992. What Microsoft calls lazy write is also
called
write-back caching
,
write-behind caching
, and
deferred
or
delayed writes
.
In modern versions of Windows, this is controlled by the Windows cache manager. The cache
manager controls Windows-based write caching and flushes the cache quite often to minimize loss. In
addition, any application that uses write operations can request that those operations be done as
write-through
, which bypasses the lazy writer and goes straight to the disk. The truth is, to minimize
the potential for lost data, lazy write (Windows write caching) is disabled for
removable
media in
Windows by default. This was designed to minimize problems stemming from what Microsoft calls
the
surprise removal
of certain hardware. Lazy write is enabled by default for internal (ATA or
SATA) storage devices inside the computer that cannot be surprise removed. Lazy write is also
enabled for certain high-performance external storage devices that might or might not be hot-
swappable, such as those attached via eSATA.
Many people worry about write caching as related to potential data loss. A worst-case scenario is a
power outage while saving a file, which will certainly cause the loss of data that hasn't already been
