How can Fedora determine the maximum speed of network computer cards?

Rick Stevens ricks at alldigital.com
Fri Nov 13 18:06:45 UTC 2015


On 11/13/2015 06:27 AM, Doug H. wrote:
> On Fri, 2015-11-13 at 13:25 +0000, Paul Smith wrote:
>> Dear All,
>>
>> Is there something in Fedora that I might use to determine the
>> maximum
>> speed the network card of my computer can attain on Internet? I am
>> thinking about wired Internet and not about wireless Internet.
>>
>> Thanks in advance,
>
> You might need to explain the reason for your request since, it may
> change the answer.
>
> As an example, my computer has a GigE port that runs to a 100Mbit
> simple switch before getting to anything else.  So I would be limited
> to a theoretical maximum of 100Mbit but I might want to know how fast I
> can transfer files to/from another computer in my LAN.  The bottleneck
> here could be the switch, my NIC, the other computers NIC, OS limits,
> firewall slowdowns etc.
>
> So, for me I would really want to test an actual file transfer using a
> number of setups.  Depending on each computer I could try Samba, FTP,
> SCP.  I might run a wire to bypass the switch.  I might try with
> firewalls and antivirus disabled.
>
> All this will also give me some clue on how fast I could run if I was
> able to get a fiber connection, to let me know if my system would be a
> serious limit to getting the full 1G of true fiber.  Obviously the
> switch would have to be upgraded in my case.

Generally speaking, any given NIC that has proper firmware and is in
a relatively modern computer will run very close to the rated wire
speed--provided the other end of the cable can handle it as well. Note
that the effective transfer rate will usually be about 80-95% of that
(for a 1Gbps NIC, 800-950Mbps).

This does not include any overhead involved in the protocol used. You
will find that things like FTP, rsync and the like will show lower
data transmission rates (e.g. multiplying the number of bytes
transferred by 8 to get number of bits transferred) due to their
overhead and that they're dealing with TCP's inherent nagle algorithm
and handshaking. TCP is designed to make sure things get to where
they're supposed to go, and with that comes a lot of overhead. Things
like UDR (rsync-over-UDP) will show much closer to the theoretical data
transmission rate due to its use of UDP.

The most common bottleneck to raw speed is the wire, switch, router,
gateway or ISP that the NIC is connected to. You'll rarely find the NIC
itself the limiting factor.
----------------------------------------------------------------------
- Rick Stevens, Systems Engineer, AllDigital    ricks at alldigital.com -
- AIM/Skype: therps2        ICQ: 226437340           Yahoo: origrps2 -
-                                                                    -
-   NEWS FLASH! Intelligence of mankind decreasing!  Details at...   -
-     uh, when, uh, the little hand is, uh, on the...  Aw, NUTS!     -
----------------------------------------------------------------------


More information about the users mailing list