[nSLUG] Slow file transfer on my LAN

George N. White III gnwiii at gmail.com
Sun Nov 1 15:27:05 AST 2009


On Sun, Nov 1, 2009 at 3:12 PM, Hatem Nassrat <hnassrat at gmail.com> wrote:
> Oliver,
>
> On Sun, Nov 1, 2009 at 1:38 PM, odoepner <odoepner at gmail.com> wrote:
>> odoepner wrote:
>>> File transfers in my home LAN never seem to be faster than 2MB/s.
>
> 2MB/s is == 16Mbps
>
> since this is greater than 11Mbps and you did mention wireless so I
> guess you are using atleast 802.11g. If you have multiple machines
> connected via wireless the are space bandwidth is divided (freq
> modulation) by the number of machines connected, i.e. if you have 3.3
> machines connected via wireless that would be the max bandwidth you
> would get on an 802.11g network.
>
> [...]
>>> The file server is an Excito Bubba/2 (has a Gigabit NIC):
>>> http://www.excito.com/bubba/technical-specifications.html
>
> Sometimes the network wiring itself may contribute to the speed. I
> recently found out that to use Gigabit one should use cat6e wiring.

Well, gigabit was designed to work on cat5(e):

"Able to achieve Gigabit speeds on commonly deployed CAT-5
cabling without costly cable replacement"
<http://www.intel.com/network/connectivity/products/pro1000mt_server_adapter.htm>

Another document is more realistic"

"Testing Existing Cable
Existing Cat-5 cabling must meet
certain transmission characteristics
before it can be used for Gigabit
Ethernet. Network managers need
not be overly concerned, since it is
estimated that less than 10% of existing
Cat-5 installations will not meet the
requirements (Gigabit Ethernet
Alliance, 2000). These installations
would also not support 100BASE-TX
Fast Ethernet."

http://www.intel.com/network/connectivity/resources/doc_library/white_papers/gigabit_ethernet/gigabit_ethernet.pdf

I'd day many network managers do need to consider cable
replacement:

Intel's Win drivers for their gigabit ethernet interfaces have a cable test
function that has been very useful at work where we have early
cat5e that was marginal even for 100-baseT.  I find the cable test
just hangs on the drops that just don't work for gigabit, but
reports "good" signal quality for the drops that work.  About
25% of our drops fail the test, and they seem to be either "test
fails completely" or test reports "good".

If we have to replace 25% of the drops we light as well redo them
all.

> Hope my assumptions are not too trivial or they were covered earlier.
> I believe this topic was started before I joined as I have no
> recollection of it.

-- 
George N. White III <aa056 at chebucto.ns.ca>
Head of St. Margarets Bay, Nova Scotia



More information about the nSLUG mailing list