| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 2021.1 | | NETCAD::STEFANI | | Tue Apr 23 1996 13:27 | 16 |
| >>1. Is Full Duplex FDDI really 200Mb/s in the same way as FDDI is
>>100Mb/s. Or is this marketing hype should be be using a lower number?
FDDI Full Duplex means that there's no token being passed and that you
can effectively transmit and receive at the same time. Yes, the
theoretical maximum throughput doubles to 200Mbps. In reality, there
may be hardware constraints that will prevent you from achieving it.
In the case of the PDQ-based adapters (DEFEA, DEFPA, DEFTA, DEFQA,
DEFAA, and DEFPZ) I believe the PDQ DMA engine can handle up to
160Mbps, so there is a cap.
Of course, in the case of end nodes, you'll have a difficult time today
driving close to that because of other non adapter related constraints.
- Larry
|
| 2021.2 | | 35356::RABAHY | dtn 471-5160, outside 1-810-347-5160 | Tue Apr 23 1996 16:42 | 6 |
| I think there must be an additional benefit to full duplex; packets are not
forwarded around the ring to be removed by the originator. In a shared ring, I
wonder how much bandwidth is lost to this?
As is often the case, if data is only flowing in one direction then, naturally,
you're only going to get a maximum of 100Mb/s.
|
| 2021.3 | | STEVMS::PETTENGILL | mulp | Fri Apr 26 1996 03:15 | 12 |
| >As is often the case, if data is only flowing in one direction then, naturally,
>you're only going to get a maximum of 100Mb/s.
Not true. While data is flowing in one direction, control is flowing in the
other, and with FDX, the control is not delayed by the data traffic. Its
really tricky to characterize in the real world, but I think that we showed
that with the DEMFA and a 3 nodes transmitting to 1, the thruput was higher
with FDX than simply removing the control traffic would gain you. This was
with the DEMFA and we found a fifo overflow problem which had to be isolated
and fixed, so I think the characterization didn't get completed and written up.
(I think that there were also NISCS protocol/implementation tuning going on
as well; NISCS is almost certainly the state of the art.)
|
| 2021.4 | | 35356::RABAHY | dtn 471-5160, outside 1-810-347-5160 | Fri Apr 26 1996 12:07 | 6 |
| re .3:
Your point is valid. However, even though the control traffic is unhindered,
you'll only ever get 100Mb/s through. Yes, this is faster than half duplex.
The only way to get more than 100Mb/s is to total both directions.
|