The standard answer is 12KB, but I believe that this statement is based on an
engineering requirement that the maximum allowed probability that the CRC fails
to detect a corrupted packet is below some threshold. (I would greatly
appreciate it if somebody can provide a specific reference for this
calculation.)
The logic behind the calculation seems to be flawed in the following way: If
you have a large quantity of data to move (say a Peta Byte, 1e+15 Bytes) then
the total undetected error rate is independent of the packet size across a huge
range of sizes. This is because changes in per-packet exposure to undetected
errors are exactly offset by the change in the number of packets.
For example if I send 1e+15 bytes over a link that has a raw error rate of
1E-12 then I compute:
Using 1000 byte packets:
Raw bit error rate: 1.0e-12
Per packet error rate: 1.0e-09
Undetected packet errors: 2.5e-19
Packets per data set: 1.0e+12
Total undetected errors per data set: 2.5e-07
Using 10000000 byte packets:
Raw bit error rate: 1.0e-12
Per packet error rate: 1.0e-05
Undetected packet errors: 2.5e-15
Packets per data set: 1.0e+08
Total undetected errors per data set: 2.5e-07
You can find some of my musing about the Internet cell size at
http://www.psc.edu/~mathis/MTU/
Comments and feedback appreciated,
--MM--
On Mon, 6 May 2002, Dr G Fairhurst wrote:
>
> I was asked by someone how big a frame size can a CRC-32 reasonably protect?
>
> One of my students came across:
>
> http://sd.wareonearth.com/~phil/jumbo.html
>
> which seems to suggest that the CRC-32 is good for 10-12KB or so of
> data, and that
> this could be one reason why we don't see many subnetworks supporting
> MTUs > 12KB.
>
> Is there more wisdom out there on the use of CRCs over large subnetwork PDUs?
> - perhaps the ATM or gigEthernet people have thought this one through?
>
>
> Gorry Fairhurst.
>
-- --MM-- ---------------------------------------------- Matt Mathis <mathis@psc.edu> W:412.268.3319 http://www.psc.edu/~mathis H:412.654.7529 ----------------------------------------------_______________________________________________ pilc mailing list pilc@ietf.org https://www1.ietf.org/mailman/listinfo/pilc
This archive was generated by hypermail 2b29 : Tue May 07 2002 - 11:35:40 EDT