mapped drive traffic route

[DBruceM2]

Trying to understand the path in which data flows when transfering data
between 2 mapped drives accross a 10/100 network. I've been reluctant
to find a straitforward answer anywhere
..
Scenario: Windows XP workstation has 2 mapped drives. Each drive is
located on a different server. All are connected via 10/100. (3 PCs
total, 1 workstation, 2 servers)
If I copy/paste or move files from one mapped drive to the other, what
path does the data take? Moreover, does the file get moved directly
from server A(mapped drive #1) to server B(maped drive #2) OR does the
file temporarily get moved from server A to workstation and then back
through the network to serverB?

My rudimentary testing indicates the 2nd scenario, since I am seeing a
significant decrease in overall transfer rate. Essentially, I am
trying to move files from server A to server B w/o logging into either
one locally and minimize network traffic at the same time. Opionions,
thoughts, ideas are welcome.

Thank you for reading.
David

 

[Bob Willard]

Trying to understand the path in which data flows when transfering data
between 2 mapped drives accross a 10/100 network. I've been reluctant
to find a straitforward answer anywhere
.
Scenario: Windows XP workstation has 2 mapped drives. Each drive is
located on a different server. All are connected via 10/100. (3 PCs
total, 1 workstation, 2 servers)
If I copy/paste or move files from one mapped drive to the other, what
path does the data take? Moreover, does the file get moved directly
from server A(mapped drive #1) to server B(maped drive #2) OR does the
file temporarily get moved from server A to workstation and then back
through the network to serverB?

My rudimentary testing indicates the 2nd scenario, since I am seeing a
significant decrease in overall transfer rate. Essentially, I am
trying to move files from server A to server B w/o logging into either
one locally and minimize network traffic at the same time. Opionions,
thoughts, ideas are welcome.

Thank you for reading.
David

Copy is a pretty dumb process. It reads a chunk of the source file into
the local PC's RAM, then writes that chunk from that RAM into the target
file, then it repeats until EndOfFile occurs; and it works the same way
whether the source file is on a local HD or a mapped HD; and it works the
same way whether target file is on a local HD or a mapped HD. (Copy may
do some overlapped I/O to speed things up, but that doesn't change the
overall data flow.)

Windows could take the shortcut path you hoped for, but that method
(called third-party I/O) gets rather complicated when errors occur.
So, M$ takes the simple approach, as outlined above.