Gigabit speed

[Riccardo]

I have two kind of Bell laptops with both Gigabit NIC which are linked
to AT-GS924 (unmanaged switch with 24 ports in Gigabit).
I copied a file of 1 GByte from laptop A to laptop B it twice or 3
times and transfering time has been about among 35 and 60 sec. , while
for my opinion it must be about 10 seconds.
Both laptops in local network icon (Windows XP Pro) show link is 1
Gbit.

If I link the 2 laptops with cross ethernet cable I have similar time.
what do you think ?

 

[JR Weiss]

I have two kind of Bell laptops with both Gigabit NIC which are linked
to AT-GS924 (unmanaged switch with 24 ports in Gigabit).
I copied a file of 1 GByte from laptop A to laptop B it twice or 3
times and transfering time has been about among 35 and 60 sec. , while
for my opinion it must be about 10 seconds.

If I link the 2 laptops with cross ethernet cable I have similar time.
what do you think ?

About right...

I get about 300 Mbps average on large files to/from my laptop. I suspect the HD
on the laptop is the choke point.

 

[Riccardo]

When I did performance testing with my computers, I used a
"ramdisk" for the data source and data destination. The
physical hard drive may be limiting your performance.

If you want another test environment to use, you could
try something like a Knoppix LiveCD, as the home
directory in that case is a RAM disk. And if you need
a large file to test with, the "dd" command can
be used to make one - dd if=/dev/zero of=some_file_name bs=1K count=...
Then you could experiment with SAMBA or FTP or whatever,
for your transfer test.

If your laptop has SATA, maybe something like this would
work. It is a SATA flash drive that can sustain 207.7MB/sec
write speed. Which is enough to use a Gigabit link to capacity.

http://techreport.com/articles.x/15931

    Paul

Hard disk of my two laptop have SATA interface (it's no specified if
sata1 or sata2).
It's good idea to test with ramdisk.
I noted this one:

DOWNLOAD test:
Laptop A: I download file (1 GB) from laptop B 35 seconds
Laptop B: I download file (1 GB) from laptop A 30 seconds

UPLOAD TEST:
Laptop A: I upload file (1 GB) from laptop B 52 seconds
Laptop B: I upload file (1 GB) from laptop A 53 seconds

1- Why upload is more slow then download operation ?
2- If My laptops had sata1 hd controller (150 MB/s) , transfer time
should have to be about 7 sec for 1 GB file ?!

 

[Paul]

Hard disk of my two laptop have SATA interface (it's no specified if
sata1 or sata2).
It's good idea to test with ramdisk.
I noted this one:

DOWNLOAD test:
Laptop A: I download file (1 GB) from laptop B 35 seconds
Laptop B: I download file (1 GB) from laptop A 30 seconds

UPLOAD TEST:
Laptop A: I upload file (1 GB) from laptop B 52 seconds
Laptop B: I upload file (1 GB) from laptop A 53 seconds

1- Why upload is more slow then download operation ?
2- If My laptops had sata1 hd controller (150 MB/s) , transfer time
should have to be about 7 sec for 1 GB file ?!

Regarding point 2, the interface to the controller on the drive
runs at 150MB/sec. But the interface to the media (the thing that
sends data to the heads), runs at a lower speed. That is the
"sustained" transfer rate. On an older drive, that is 60MB/sec.
On a modern desktop drive, it can be 90MB/sec. On a
Velociraptor it is 120MB/sec. The flash drive mentioned, is 200MB/sec.

To test your hard drive transfer rate curve, run this program. It
will show the sustained transfer rate, as a function of the position
of the head on the platter surface.

Select version 2.55, as that is free, and only does a read benchmark.
Being a read benchmark, that will not hurt your hard drive.

http://www.hdtune.com/download.html

The reason I did not recommend a RAM Disk program to you, is none
of the ones I played with were that good. Some of the RAM Disk code
used, is based on some code released by Microsoft as an example. So
some of the programs share a common origin.

I may have used this for my testing. Usually, RAM Disk programs
have some limitations, as to how much RAM they can use. This may be
a different version than the one I used (as I did my test at least
a couple years ago).

http://www.arsoft-online.com/index....&id=26:ar-ram-disk&catid=25:ramdisk&Itemid=47

There is another option here. This may still be under development.

http://www.cenatek.com/product_page_ramdisk_download_list.php

There are a total of four tests you can try. You can read or write
from either side of the pair of computers.

Be careful when dealing with the files, as the operating system
file cache may be involved. You need a way to nullify the file
cache, so you don't get a cached copy of the file instead of a
fresh copy. This may involve some thinking about ways to defeat it,
if present. For example, if you copy the same file over the link
twice, the second copy could be much faster, due to local caching.

With some luck, you may get to see something closer to the full
link rate.

HTH,
Paul

 

[Riccardo]

Regarding point 2, the interface to the controller on the drive
runs at 150MB/sec. But the interface to the media (the thing that
sends data to the heads), runs at a lower speed. That is the
"sustained" transfer rate. On an older drive, that is 60MB/sec.
On a modern desktop drive, it can be 90MB/sec. On a
Velociraptor it is 120MB/sec. The flash drive mentioned, is 200MB/sec.

To test your hard drive transfer rate curve, run this program. It
will show the sustained transfer rate, as a function of the position
of the head on the platter surface.

Select version 2.55, as that is free, and only does a read benchmark.
Being a read benchmark, that will not hurt your hard drive.

http://www.hdtune.com/download.html

The reason I did not recommend a RAM Disk program to you, is none
of the ones I played with were that good. Some of the RAM Disk code
used, is based on some code released by Microsoft as an example. So
some of the programs share a common origin.

I may have used this for my testing. Usually, RAM Disk programs
have some limitations, as to how much RAM they can use. This may be
a different version than the one I used (as I did my test at least
a couple years ago).

http://www.arsoft-online.com/index.php?option=com_content&view=articl....

There is another option here. This may still be under development.

http://www.cenatek.com/product_page_ramdisk_download_list.php

There are a total of four tests you can try. You can read or write
from either side of the pair of computers.

Be careful when dealing with the files, as the operating system
file cache may be involved. You need a way to nullify the file
cache, so you don't get a cached copy of the file instead of a
fresh copy. This may involve some thinking about ways to defeat it,
if present. For example, if you copy the same file over the link
twice, the second copy could be much faster, due to local caching.

With some luck, you may get to see something closer to the full
link rate.

HTH,
      Paul

Thank you very much for your precious suggestions which I'll follow as
soon as possible.

 

[Riccardo]

Regarding point 2, the interface to the controller on the drive
runs at 150MB/sec. But the interface to the media (the thing that
sends data to the heads), runs at a lower speed. That is the
"sustained" transfer rate. On an older drive, that is 60MB/sec.
On a modern desktop drive, it can be 90MB/sec. On a
Velociraptor it is 120MB/sec. The flash drive mentioned, is 200MB/sec.

To test your hard drive transfer rate curve, run this program. It
will show the sustained transfer rate, as a function of the position
of the head on the platter surface.

Select version 2.55, as that is free, and only does a read benchmark.
Being a read benchmark, that will not hurt your hard drive.

http://www.hdtune.com/download.html

The reason I did not recommend a RAM Disk program to you, is none
of the ones I played with were that good. Some of the RAM Disk code
used, is based on some code released by Microsoft as an example. So
some of the programs share a common origin.

I may have used this for my testing. Usually, RAM Disk programs
have some limitations, as to how much RAM they can use. This may be
a different version than the one I used (as I did my test at least
a couple years ago).

http://www.arsoft-online.com/index.php?option=com_content&view=articl....

There is another option here. This may still be under development.

http://www.cenatek.com/product_page_ramdisk_download_list.php

There are a total of four tests you can try. You can read or write
from either side of the pair of computers.

Be careful when dealing with the files, as the operating system
file cache may be involved. You need a way to nullify the file
cache, so you don't get a cached copy of the file instead of a
fresh copy. This may involve some thinking about ways to defeat it,
if present. For example, if you copy the same file over the link
twice, the second copy could be much faster, due to local caching.

With some luck, you may get to see something closer to the full
link rate.

HTH,
      Paul

And what do you think about point 1 ? Upload is more slow then
download operation.

bye

 

[Paul]

And what do you think about point 1 ? Upload is more slow then
download operation.

bye

The computer has multiple subsystems. There is the network
part. There is the storage part. The two have to be
coordinated. The storage part may choose to "read-ahead"
or "write-cache" for example. Perhaps there are
differences in the two scenarios, as to whether the OS
is willing to do those things or not (pipeline the operations).

Paul