At 02:33 PM 1/23/01, Brian Lodwick wrote:
> I think the only question that is left unanswered is, if anyone bothered
>to do the math, why does the minimum frame size only increase by a multiple
>of 8 and not 10?
>512 / 64 = 8
Hi Brian,
I hope our conversation isn't like the reserved token on FDDI and we're the
only ones still listening! &;-)
If the IEEE had just been changing the physical diameter, a factor of 10
would have made more sense (especially since wiring specifications are
often in metric). Since they were also changing the number of bits that the
MAC sends, they probably wanted that to be in octets. (That is just a
guess! &;-)
Here's another question. How badly did they affect efficiency by
"enhancing" the CSMA/CD this way?
Does a receiver throw out a frame if the collision happened in the carrier
extension part even though the actual data part arrived without error?
(Yes! It has to because the sender is going to resend and Ethernet can't
deal with duplicate frames.)
And how about all that overhead for small frames? Let's say the next layer
up sends a 512-bit frame. That ends up being a 4096-bit frame with a 64-bit
preamble and 96-bit interframe gap.
Consider an application that is trying to output a bunch of small frames at
a quick rate. If it has to add a bunch of bits to reach 4096, will it be
able to send at the proper rate? When it's done and releases control, some
other application will probably jump in, and the poor small-frame
application has to try to get control again.
Well, they addressed this also. The IEEE standard lets a sender send
multiple frames! According to Rich Seifert in his "Gigabit Ethernet" book,
a station may choose to burst frames if there is a frame in its transmit
queue when it has finished sending its first frame plus extension (if
necessary). It does this without contending again for use of the channel.
During the interframe gap it sends non-data symbols and then sends its next
frame.
But is that fair to other senders? Well, was Ethernet ever fair? It was
fair from an access point of view, but senders of big frames could use a
greater portion of the bandwidth. This new bursting feature makes it fair
from a bandwidth usage point of view. A station can send for up to a
burstLength time (plus one frame) regardless of whether the frames are
short or long. This is a major difference. It's like custom queuing versus
priority queuing.
I'm beginning to think that Gigabit Ethernet is best deployed on a
full-duplex (not shared) link. I get squeamish when I hear about all these
changes to the CSMA/CD that we have known and loved for so long. &;-)
Priscilla
> >>>Brian
>
>* IEEE article can be found at: http://computer.org/Internet/v1n5/ether.htm
>
> >From: Priscilla Oppenheimer <[EMAIL PROTECTED]>
> >To: "Brian Lodwick" <[EMAIL PROTECTED]>,[EMAIL PROTECTED],
> >[EMAIL PROTECTED]
> >Subject: Re: EASY ?? FOR MOST OF YOU
> >Date: Mon, 22 Jan 2001 16:15:31 -0800
> >
> >At 09:20 PM 1/22/01, Brian Lodwick wrote:
> >>BBBBBBBBBBBBBBBBBBBBBBBBBBBBB for
> >>
> >> >>>Brian
> >>My additional question on top of this one is:
> >>
> >>If the maximum legal length was set to 1/10th the size to make regular
> >>(10Mb/sec) increased by a factor of 10, what was done to further increase
> >>100Mb/sec Ethernet by a factor of 10 to get Gigabit Ethernet?
> >
> >I like your plan to turn this into a non-easy question! The only experience
> >I have with Gigabit Ethernet is in a fully-switched network where every
> >port is full duplex, in which case CSMA/CD parameters are not an issue.
> >However, shared, half-duplex Gigabit Ethernet is viable also.
> >
> >With shared 10 and 100-Mbps Ethernet, the minimum frame size is equal to
> >the maximum round-trip propagation delay of the network. In other words,
> >the minimum frame size is equal to the slotTime = 512 bits. Sticking to
> >this rule would haver resulted in impracticably small networks for Gigabit
> >Ethernet, however. The solution was a process called "carrier extension."
> >
> >According to Rich Seifert in his excellent book, "Gigabit Ethernet," "The
> >key change is that the slotTime and the minimum frame are no longer the
> >same. The minimum frame is maintained at 512 bits (64 bytes, as in 10 Mbps
> >and 100 Mbps Ethernets), but the slotTime is set at 4096 bit-times (512
> >bytes).
> >
> >Frames that are shorter than the slotTime are artificially extended by
> >appending a carrier-extension field so that they are exactly one slotTime
> >long. This extends the duration of the time that the station transmits....
> >If a collision occurs during any time from the beginning of the frame to
> >the end of the extension field, the MAC will jam, abort, and backoff."
> >
> >See the book for even more gory details! &;-)
> >
> >Priscilla
> >
> >
> >
> >> >>>Brian
> >>
> >>
> >>attenuation is effected by 3 elements spreading, scattering, and
> >>absorption.
> >>
> >> >From: Alvarado Jesus <[EMAIL PROTECTED]>
> >> >Reply-To: Alvarado Jesus <[EMAIL PROTECTED]>
> >> >To: "'[EMAIL PROTECTED]'" <[EMAIL PROTECTED]>
> >> >Subject: EASY ?? FOR MOST OF YOU
> >> >Date: Mon, 22 Jan 2001 14:10:43 -0500
> >> >
> >> >The network span of a 100Base-T Network (205) meters is approx. 10 times
> >> >smaller than the network span of a 10Base-T network (2500) meters
> >>Because
> >> >....
> >> >
> >> >A) , Higher speed data signals attenuate more quickly and so cannot be
> >> >transmitted that far
> >> >
> >> >
> >> >B) . Both Networks have minimum frame sizes of 64 bytes and the network
> >> >spans must be tied directly to the minimum frame transmission time to
> >>avoid
> >> >collisions.
> >> >
> >> >_________________________________
> >> >FAQ, list archives, and subscription info:
> >> >http://www.groupstudy.com/list/cisco.html
> >> >Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
> >>
> >>_________________________________________________________________
> >>Get your FREE download of MSN Explorer at http://explorer.msn.com
> >>
> >>_________________________________
> >>FAQ, list archives, and subscription info:
> >>http://www.groupstudy.com/list/cisco.html
> >>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
> >
> >
> >________________________
> >
> >Priscilla Oppenheimer
> >http://www.priscilla.com
> >
>
>_________________________________________________________________
>Get your FREE download of MSN Explorer at http://explorer.msn.com
>
>_________________________________
>FAQ, list archives, and subscription info:
>http://www.groupstudy.com/list/cisco.html
>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
________________________
Priscilla Oppenheimer
http://www.priscilla.com
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
