Joe Abley wrote:
On Fri, Feb 23, 2001 at 12:02:05PM +1300, Stephen Donnelly wrote:
The 'cell tax' is 5 bytes of overhead for every 48 bytes of data, or about 1/10. So for OC12 (around 622Mbps), it's about 60Mbps, not 155, and for oc48 it's around 250Mbps. That is over an OC3, but much less than an OC12.
The 5 bytes of header per cell is part of the cell tax; there's also the wastage caused by (a) AAL5 and possibly SNAP encapsulation, and (b) needing to populate an integer number of cells when the total frame-based payload is not a multiple of 48 bytes.
Worst case, you will carry 49 bytes of payload per two cells. That's 57 bytes wastage in 106 bytes transmitted, or a cell tax of around 54%.
The extent to which this effect is significant depends on the distribution of packet sizes in your traffic. Unfortunately, on the internet, a high proportion of packets are TCP ACKs with no data, which narrowly fail to fit inside a single cell.
On the internet, a cell tax of 30% or so is quite normal.
You're right there, of course. I'm still in the habit of thinking of raw cell payloads as opposed to actual useful bytes of IP in AAL5 etc. Ethernet also only gets about 50% efficiency for small packets though, if you include the IFG. Cisco POS does better, with only 4 bytes header per packet (plus a start byte and 2 to end), but I think there's a minimum frame size there too in practice that would lower the efficiency slightly. POS also has to do byte stuffing, as it's basically HDLC. Stephen. -- ----------------------------------------------------------------------- Stephen Donnelly (BCMS) email: sfd(a)cs.waikato.ac.nz WAND Group Room GG.15 phone +64 7 838 4086 Computer Science Department, University of Waikato, New Zealand ----------------------------------------------------------------------- --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog