100Mbps Ethernet Speed/Efficiency
Rodolfo J. Paiz
rpaiz at simpaticus.com
Tue Apr 6 14:19:49 UTC 2004
At 04:49 4/6/2004, you wrote:
> >From a network standpoint the following applies:
>
>100M(Bits)ps = 100,000,000Bps
>Full Duplex = 200,000,000Bps
>In Bytes = 100,000,000/8 = 12,500,000
>In bytes FD = 200,000,000/8 = 25,000,000
>Typical Data link + TCP/IP net Overhead = 30%
>Typical maximum theoretical throughput then is 70% of available
>bandwidth
FYI, a few things which come in handy:
1. In communications, everything is in powers of 10. So K = 1,000,
M = 1,000,000, and 56 Kbps = 56,000. In computing, however, everything is
in powers of 2, so K = 1,024, and M = 1,048,576.
2. A bit should be represented by a lowercase "b" and a byte as an
uppercase "B". So the notation above is incorrect... 100 Mbps should have a
"b" for bits.
3. Metric prefixes like K, M, and G should always be capitalized.
And yes, my 100 Mbps switches (full-duplex) generally provide for a one-way
file transfer right around 60-70 Mbps which seems to be near the useful
maximum of the network. Transfer speeds below that should be host-related,
and transfer speeds much above 70 Mpbs are likely a mistake. :-)
>Ow Mun Heng wrote:
>
> >Transferring files between 2 PCs, (laptop and Desktop)
> >I see like up to 20MB/s.
> >
> >Could this # be limited due to my slow HD? 4200rpm which hdparm -t (or
> >is it -T) gives ~26MB/s
This shows a classic mistake in labeling. Transfers are supposed to be
20MBps which would be 160 Mbps which is not bloody likely, whereas hdparm
output could quite realistically be 26MBps (~208 Mbps). I can believe the
hard drive output, but not the transfer speeds. Most likely Ow Mun Heng is
getting 20 Mbps per second which looks slow to me. But we can't be sure
without more data from him.
----
Following this thread and keeping responses in order sure would be a lot
easier if you guys would post answers and follow-ups at the bottom, instead
of at the top, guys. Top-posting is great for one-on-one communication, but
badly breaks the chronological and sequential flow of a discussion among a
group like a mailing list. That's why convention on mailing lists and
newsgroups is to trim the text of previous messages, leaving only what is
relevant for context, then reply below the quoted text.
More work for the sender? Yes. But when you're in a group of 4,000 people,
courtesy demands that you spend a moment or two to make reading the message
easier for the other 4,000 people, and self-interest suggests that those
people who send properly-formatted and coherent messages are likely to have
their messages read and answered by more members of the community. When
people can read your stuff more easily, you are more likely to get help.
--
Rodolfo J. Paiz
rpaiz at simpaticus.com
http://www.simpaticus.com
More information about the fedora-list
mailing list